{"llms":[{"_id":"6acd923fd5a03c8c42c015b3","name":"Claude Haiku 5.5","slug":"claude-haiku-5-5","apiIdentifier":"claude-haiku-5-5","family":"Claude","version":"5.5","description":"Claude Haiku 5.5 is Anthropic's fast proprietary model for high-volume classification, extraction, routing, and subagent workloads. It accepts text and images, returns text, uses adaptive thinking, and supports a 1,000,000-token context window with up to 128,000 output tokens on standard API calls.","organizationSlug":"anthropic","organizationName":"Anthropic","contextWindow":1000000,"maxOutputTokens":128000,"inputPricePerMillion":0.1,"outputPricePerMillion":0.5,"capabilities":["reasoning","vision","tool-use","long-context","context-caching"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-10-07T00:00:00.000Z","tags":["reasoning","vision","long-context","api","claude","fast"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3.2","score":43.3950199670746,"canonicalName":"Artificial Analysis Intelligence Index v4.3.2","rawName":"Artificial Analysis Intelligence Index v4.3.2","sourceUrl":"https://artificialanalysis.ai/models/claude-haiku-5-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3.2","modelVariant":"Claude Haiku 5.5 (Max)","exactModelVariant":"Claude Haiku 5.5 (Max)","configuration":{"protocol":"aa-independent-2026-09-v4.3.2","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","components":["AA-Briefcase v1.1","GDPval-AA v2.1","AutomationBench-AA","Terminal-Bench 4.0","SciCode","Humanity's Last Exam","GDP.pdf","CritPt","AA-Omniscience","AA-LCR v1.1"],"modelSettings":{"evaluatorModelId":"7d064a11-6e78-4030-b535-cdf940615041","evaluatorModelSlug":"claude-haiku-5-5","reasoning":true,"effort":"max","selection":"Exact non-estimated evaluator row; one API model, not a separate max-effort product"}},"benchmarkVersion":"4.3.2","artifactId":"aa-claude-haiku-5-5-cb3da42621aaf0cd","evidenceId":"claude-haiku-5-5:aa:intelligenceIndex:v4.3.2:max","confidence":"high","higherIsBetter":true,"comparisonEligible":true,"estimated":false,"date":null,"observedAt":"2026-10-13T02:02:12.781Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6ac328a3e1f3c43d16b11889","name":"Strands Decider","slug":"strands-decider","apiIdentifier":"StrandsAgents/strands-decider-2B-hobson-v19","family":"Strands Decider","version":"2B Hobson v19","description":"Strands Decider v19 is an Apache-2.0, open-weight 1.9B decision model for agentic workflows. It selects among options or rates inputs and returns calibrated confidence scores without generating text.","organizationSlug":"amazon","organizationName":"Amazon","contextWindow":4096,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","classification","scoring","routing","confidence"],"category":"Decision Model","current":true,"deprecated":false,"releaseDate":"2026-10-05T04:33:36.556Z","tags":["decision-model","agentic-ai","classification","confidence","routing","open-source"],"benchmarks":[{"source":"official-readme","name":"JevBench public-set accuracy","score":72.3,"canonicalName":"JevBench public-set accuracy","rawName":"JevBench v1 public set, 231 tasks, accuracy","sourceUrl":"https://raw.githubusercontent.com/strands-labs/strands-decider/main/evaluation/results.md","sourceType":"first-party-model-evaluation","unit":"percent","scale":"0-100","modelVariant":"Strands Decider 2B Hobson v19","configuration":{"protocol":"jevbench-public-v1.4.2-snapshot-2026-09-25","harness":"JevBench official harness at pinned commit","subset":"231 public tasks","metric":"accuracy","evaluationWindowTokens":3072,"servedDefaultWindowTokens":4096,"resultCount":"167/231","modelSettings":{"checkpoint":"StrandsAgents/strands-decider-2B-hobson-v19","recipe":"v19","calibrated":true}},"benchmarkVersion":"JevBench v1 public / board v1.4.2 snapshot 2026-09-25","artifactId":"strands-evaluation-results","evidenceId":"strands-decider:jevbench:v19:3072:2026-10-05","confidence":"high","higherIsBetter":true,"comparisonEligible":false,"exclusionReason":"Source-reported JevBench result; not an independently observed globally selected comparison protocol.","date":"2026-10-05T04:33:36.556Z","observedAt":"2026-10-05T14:16:52.945Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6abbfda06a1c2a13703b7f75","name":"GPT-6.1 Sol","slug":"gpt-6-1-sol","apiIdentifier":"gpt-6.1-sol","family":"GPT-6.1","version":"Sol","description":"GPT-6.1 Sol is OpenAI's proprietary reasoning model for coding, computer use, and professional workflows. The gpt-6.1-sol API model accepts text and images and returns text. OpenAI documents a 1,050,000-token context window, 922,000 maximum input tokens, and 128,000 maximum output tokens. OpenAI reports first-party HealthBench results; Artificial Analysis independently evaluated the max-effort configuration under its v4.3.2 Intelligence Index.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":2,"outputPricePerMillion":10,"capabilities":["text","vision","reasoning","coding","tool-use","computer-use","long-context","context-caching"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-09-29T00:00:00.000Z","tags":["reasoning","coding","computer-use","agentic","vision","api","long-context"],"benchmarks":[{"source":"openai","name":"HealthBench Professional length-adjusted","score":64.2,"canonicalName":"HealthBench Professional length-adjusted","rawName":"HealthBench Professional length-adjusted","sourceUrl":"https://cdn.openai.com/pdf/38e3efcf-545e-44cd-99ec-2b7eb395f4cc/oai_GPT_6_1_Sol.pdf","sourceType":"first-party-system-card","unit":"percent","scale":"0-100 length-adjusted score","modelVariant":"GPT-6.1 Sol","exactModelVariant":"GPT-6.1 Sol","configuration":{"protocol":"openai-gpt-6-1-sol-system-card-table-7","metric":"length-adjusted score","unadjustedScore":67.2,"meanResponseCharacters":3038,"reasoningEffort":"not specified in Table 7"},"artifactId":"openai-sol61-card-2287b79276b521b6","evidenceId":"gpt-6-1-sol:healthbench-professional:length-adjusted:card-table-7","confidence":"high","higherIsBetter":true,"comparisonEligible":false,"estimated":false,"date":null,"observedAt":"2026-09-29T17:49:54.190Z"},{"source":"openai","name":"HealthBench length-adjusted","score":58.5,"canonicalName":"HealthBench length-adjusted","rawName":"HealthBench length-adjusted","sourceUrl":"https://cdn.openai.com/pdf/38e3efcf-545e-44cd-99ec-2b7eb395f4cc/oai_GPT_6_1_Sol.pdf","sourceType":"first-party-system-card","unit":"percent","scale":"0-100 length-adjusted score","modelVariant":"GPT-6.1 Sol","exactModelVariant":"GPT-6.1 Sol","configuration":{"protocol":"openai-gpt-6-1-sol-system-card-table-7","metric":"length-adjusted score","unadjustedScore":56.7,"meanResponseCharacters":1701,"reasoningEffort":"not specified in Table 7"},"artifactId":"openai-sol61-card-2287b79276b521b6","evidenceId":"gpt-6-1-sol:healthbench:length-adjusted:card-table-7","confidence":"high","higherIsBetter":true,"comparisonEligible":false,"estimated":false,"date":null,"observedAt":"2026-09-29T17:49:54.190Z"},{"source":"openai","name":"HealthBench Hard length-adjusted","score":36.2,"canonicalName":"HealthBench Hard length-adjusted","rawName":"HealthBench Hard length-adjusted","sourceUrl":"https://cdn.openai.com/pdf/38e3efcf-545e-44cd-99ec-2b7eb395f4cc/oai_GPT_6_1_Sol.pdf","sourceType":"first-party-system-card","unit":"percent","scale":"0-100 length-adjusted score","modelVariant":"GPT-6.1 Sol","exactModelVariant":"GPT-6.1 Sol","configuration":{"protocol":"openai-gpt-6-1-sol-system-card-table-7","metric":"length-adjusted score","unadjustedScore":33.4,"meanResponseCharacters":1646,"reasoningEffort":"not specified in Table 7"},"artifactId":"openai-sol61-card-2287b79276b521b6","evidenceId":"gpt-6-1-sol:healthbench-hard:length-adjusted:card-table-7","confidence":"high","higherIsBetter":true,"comparisonEligible":false,"estimated":false,"date":null,"observedAt":"2026-09-29T17:49:54.190Z"},{"source":"openai","name":"HealthBench Consensus length-adjusted","score":96,"canonicalName":"HealthBench Consensus length-adjusted","rawName":"HealthBench Consensus length-adjusted","sourceUrl":"https://cdn.openai.com/pdf/38e3efcf-545e-44cd-99ec-2b7eb395f4cc/oai_GPT_6_1_Sol.pdf","sourceType":"first-party-system-card","unit":"percent","scale":"0-100 length-adjusted score","modelVariant":"GPT-6.1 Sol","exactModelVariant":"GPT-6.1 Sol","configuration":{"protocol":"openai-gpt-6-1-sol-system-card-table-7","metric":"length-adjusted score","unadjustedScore":95.9,"meanResponseCharacters":1686,"reasoningEffort":"not specified in Table 7"},"artifactId":"openai-sol61-card-2287b79276b521b6","evidenceId":"gpt-6-1-sol:healthbench-consensus:length-adjusted:card-table-7","confidence":"high","higherIsBetter":true,"comparisonEligible":false,"estimated":false,"date":null,"observedAt":"2026-09-29T17:49:54.190Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3.2","score":51.8332597011541,"canonicalName":"Artificial Analysis Intelligence Index v4.3.2","rawName":"Artificial Analysis Intelligence Index v4.3.2","sourceUrl":"https://artificialanalysis.ai/models/gpt-6-1-sol","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3.2","modelVariant":"GPT-6.1 Sol (max)","exactModelVariant":"GPT-6.1 Sol (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3.2","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","components":["AA-Briefcase v1.1","GDPval-AA v2.1","AutomationBench-AA","Terminal-Bench 4.0","SciCode","Humanity's Last Exam","GDP.pdf","CritPt","AA-Omniscience","AA-LCR v1.1"],"modelSettings":{"evaluatorModelId":"b25ac058-1e33-4c31-bbc9-a20c3ad0717a","evaluatorModelSlug":"gpt-6-1-sol","reasoning":true,"effort":"max","selection":"Exact non-estimated evaluator row"}},"benchmarkVersion":"4.3.2","artifactId":"aa-gpt61-sol-4d2579edcc9f8f92","evidenceId":"gpt-6-1-sol:aa:intelligenceIndex:v4.3.2:max","confidence":"high","higherIsBetter":true,"comparisonEligible":false,"exclusionReason":"BenchmarkProtocolReviewRequired: v4.3.2 is not the globally selected exact protocol","estimated":false,"date":null,"observedAt":"2026-09-30T01:33:18.930Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6abbe53e1dae6ac05d665a28","name":"Claude Sonnet 5.5","slug":"claude-sonnet-5-5","apiIdentifier":"claude-sonnet-5-5","family":"Claude Sonnet","version":"5.5","description":"Claude Sonnet 5.5 is Anthropic's proprietary adaptive-reasoning model. The Claude API identifier is claude-sonnet-5-5. It accepts text and images, returns text, has a 1,000,000-token context window and a 128,000-token standard output limit. Anthropic lists standard API input and output pricing at $2 and $10 per million tokens.","organizationSlug":"anthropic","organizationName":"Anthropic","contextWindow":1000000,"maxOutputTokens":128000,"inputPricePerMillion":2,"outputPricePerMillion":10,"capabilities":["text","vision","reasoning","coding","tool-use","long-context","context-caching"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-09-28T00:00:00.000Z","tags":["reasoning","coding","vision","long-context","api","adaptive-thinking"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3.2","score":55.9779549012591,"canonicalName":"Artificial Analysis Intelligence Index v4.3.2","rawName":"Artificial Analysis Intelligence Index v4.3.2","sourceUrl":"https://artificialanalysis.ai/models/claude-sonnet-5-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3.2","modelVariant":"Claude Sonnet 5.5 (Adaptive Reasoning, Max Effort, Default Fallback)","exactModelVariant":"Claude Sonnet 5.5 (Adaptive Reasoning, Max Effort, Default Fallback)","configuration":{"protocol":"aa-independent-2026-09-v4.3.2","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","components":["AA-Briefcase v1.1","GDPval-AA v2.1","AutomationBench-AA","Terminal-Bench 4.0","SciCode","Humanity's Last Exam","GDP.pdf","CritPt","AA-Omniscience","AA-LCR v1.1"],"modelSettings":{"evaluatorModelId":"b171d979-5ec1-45de-b8b9-db1ee65e77ec","evaluatorModelSlug":"claude-sonnet-5-5","reasoning":true,"effort":"max","fallback":"default","selection":"Exact non-estimated evaluator row"}},"benchmarkVersion":"4.3.2","artifactId":"aa-sonnet55-6c17ba1524cf07a6","evidenceId":"claude-sonnet-5-5:aa:intelligenceIndex:v4.3.2:max-default-fallback","confidence":"high","higherIsBetter":true,"comparisonEligible":true,"estimated":false,"date":null,"observedAt":"2026-09-29T16:12:23.704Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3.2","score":56.0001286580794,"canonicalName":"Artificial Analysis Intelligence Index v4.3.2","rawName":"Artificial Analysis Intelligence Index v4.3.2","sourceUrl":"https://artificialanalysis.ai/models/claude-sonnet-5-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3.2","modelVariant":"Claude Sonnet 5.5 (Adaptive Reasoning, Max Effort, Default Fallback)","exactModelVariant":"Claude Sonnet 5.5 (Adaptive Reasoning, Max Effort, Default Fallback)","configuration":{"protocol":"aa-independent-2026-09-v4.3.2","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","components":["AA-Briefcase v1.1","GDPval-AA v2.1","AutomationBench-AA","Terminal-Bench 4.0","SciCode","Humanity's Last Exam","GDP.pdf","CritPt","AA-Omniscience","AA-LCR v1.1"],"modelSettings":{"evaluatorModelId":"b171d979-5ec1-45de-b8b9-db1ee65e77ec","evaluatorModelSlug":"claude-sonnet-5-5","reasoning":true,"effort":"max","fallback":"default","selection":"Exact non-estimated evaluator row"}},"benchmarkVersion":"4.3.2","artifactId":"aa-sonnet55","evidenceId":"claude-sonnet-5-5:aa:intelligenceIndex:v4.3.2:max-default-fallback:2026-10-02","confidence":"high","higherIsBetter":true,"comparisonEligible":true,"estimated":false,"date":null,"observedAt":"2026-10-02T13:10:05.942Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6ab2d1ea588c9540a98b1b9d","name":"GPT-6 Luna","slug":"gpt-6-luna","apiIdentifier":"gpt-6-luna","family":"GPT-6","version":"Luna","description":"GPT-6 Luna is OpenAI's efficient reasoning model for focused, high-volume tasks. It accepts text and images, produces text, supports a 1,050,000-token context window and 128,000 output tokens, and is available through the Responses and Chat Completions APIs under the exact gpt-6-luna identifier.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":0.1,"outputPricePerMillion":0.5,"capabilities":["text","vision","reasoning","coding","tool-use","long-context","context-caching"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-09-22T00:00:00.000Z","tags":["reasoning","vision","long-context","api","efficient","high-volume"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3.2","score":37.2559686869738,"canonicalName":"Artificial Analysis Intelligence Index v4.3.2","rawName":"Artificial Analysis Intelligence Index v4.3.2","sourceUrl":"https://artificialanalysis.ai/models/gpt-6-luna","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3.2","modelVariant":"GPT-6 Luna (max)","exactModelVariant":"GPT-6 Luna (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3.2","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","components":["AA-Briefcase v1.1","GDPval-AA v2.1","AutomationBench-AA","Terminal-Bench 4.0","SciCode","Humanity's Last Exam","GDP.pdf","CritPt","AA-Omniscience","AA-LCR v1.1"],"modelSettings":{"evaluatorModelSlug":"gpt-6-luna","effort":"max","selection":"Exact non-estimated evaluator row"}},"benchmarkVersion":"4.3.2","artifactId":"aa-gpt-6-luna-f323f0bbe78a5cfc","evidenceId":"gpt-6-luna:aa:intelligenceIndex:v4.3.2:max","confidence":"high","higherIsBetter":true,"comparisonEligible":true,"estimated":false,"date":null,"observedAt":"2026-09-22T19:01:59.045Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6ab2d1ea588c9540a98b1b9c","name":"GPT-6 Sol","slug":"gpt-6-sol","apiIdentifier":"gpt-6-sol","family":"GPT-6","version":"Sol","description":"GPT-6 Sol is OpenAI's reasoning model for professional work. It accepts text and images, produces text, supports a 1,050,000-token context window and 128,000 output tokens, and is available through the Responses and Chat Completions APIs. OpenAI documents the exact gpt-6-sol identifier and standard token pricing.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":2,"outputPricePerMillion":10,"capabilities":["text","vision","reasoning","coding","tool-use","long-context","context-caching"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-09-22T00:00:00.000Z","tags":["reasoning","coding","vision","long-context","api","agentic"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3.2","score":47.5276426437724,"canonicalName":"Artificial Analysis Intelligence Index v4.3.2","rawName":"Artificial Analysis Intelligence Index v4.3.2","sourceUrl":"https://artificialanalysis.ai/models/gpt-6-sol","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3.2","modelVariant":"GPT-6 Sol (max)","exactModelVariant":"GPT-6 Sol (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3.2","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","components":["AA-Briefcase v1.1","GDPval-AA v2.1","AutomationBench-AA","Terminal-Bench 4.0","SciCode","Humanity's Last Exam","GDP.pdf","CritPt","AA-Omniscience","AA-LCR v1.1"],"modelSettings":{"evaluatorModelSlug":"gpt-6-sol","effort":"max","selection":"Exact non-estimated evaluator row"}},"benchmarkVersion":"4.3.2","artifactId":"aa-gpt-6-sol-d0a18fc17caf20bf","evidenceId":"gpt-6-sol:aa:intelligenceIndex:v4.3.2:max","confidence":"high","higherIsBetter":true,"comparisonEligible":true,"estimated":false,"date":null,"observedAt":"2026-09-22T19:01:50.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6ab2cd283e22c7b8722cdb8e","name":"MiMo-V2.6-Flash","slug":"mimo-v26-flash","apiIdentifier":"mimo-v2.6-flash","family":"MiMo","version":"2.6 Flash","description":"MiMo-V2.6-Flash is Xiaomi's MIT-licensed, open-weight multimodal reasoning model optimized for high-frequency and large-scale professional workloads. It supports text, image, video, and audio input, a one-million-token context window, 128K-token output, tool use, web search, structured output, and agent workflows.","organizationSlug":"xiaomi","organizationName":"Xiaomi","contextWindow":1000000,"maxOutputTokens":128000,"inputPricePerMillion":0.14,"outputPricePerMillion":0.28,"capabilities":["reasoning","multimodal","coding","tool-use","web-search","structured-output","computer-use","context-caching","open-weights","long-context"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-09-22T00:00:00.000Z","tags":["open-weights","reasoning","multimodal","coding","agents","efficient","long-context"],"benchmarks":[{"source":"xiaomi-mimo","name":"DeepSWE v1.1","score":67.9,"canonicalName":"DeepSWE v1.1","rawName":"DeepSWE v1.1","sourceUrl":"https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"MiMo-V2.6 Flash","configuration":{"protocol":"xiaomi-mimo-v2.6-model-card-2026-09-22","metric":"model-card reported score"},"benchmarkVersion":"1.1","artifactId":"xiaomi-mimo-v26-hf-7e3928885f5658a7","evidenceId":"mimo-v26-flash:model-card:deepswe-v1.1","confidence":"high","higherIsBetter":true,"date":"2026-09-22T00:00:00.000Z","observedAt":"2026-09-22T03:59:47.024Z"},{"source":"xiaomi-mimo","name":"AutomationBench v1.0.6","score":52.3,"canonicalName":"AutomationBench v1.0.6","rawName":"AutomationBench v1.0.6","sourceUrl":"https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"MiMo-V2.6 Flash","configuration":{"protocol":"xiaomi-mimo-v2.6-model-card-2026-09-22","metric":"model-card reported score"},"benchmarkVersion":"1.0.6","artifactId":"xiaomi-mimo-v26-hf-7e3928885f5658a7","evidenceId":"mimo-v26-flash:model-card:automationbench-v1.0.6","confidence":"high","higherIsBetter":true,"date":"2026-09-22T00:00:00.000Z","observedAt":"2026-09-22T03:59:47.024Z"},{"source":"xiaomi-mimo","name":"Toolathlon-Verified","score":73.6,"canonicalName":"Toolathlon-Verified","rawName":"Toolathlon-Verified","sourceUrl":"https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"MiMo-V2.6 Flash","configuration":{"protocol":"xiaomi-mimo-v2.6-model-card-2026-09-22","metric":"model-card reported score"},"artifactId":"xiaomi-mimo-v26-hf-7e3928885f5658a7","evidenceId":"mimo-v26-flash:model-card:toolathlon-verified","confidence":"high","higherIsBetter":true,"date":"2026-09-22T00:00:00.000Z","observedAt":"2026-09-22T03:59:47.024Z"},{"source":"xiaomi-mimo","name":"Terminal-Bench 4.0","score":28.8,"canonicalName":"Terminal-Bench 4.0","rawName":"Terminal Bench 4.0","sourceUrl":"https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"MiMo-V2.6 Flash","configuration":{"protocol":"xiaomi-mimo-v2.6-model-card-2026-09-22","metric":"model-card reported score"},"benchmarkVersion":"4.0","artifactId":"xiaomi-mimo-v26-hf-7e3928885f5658a7","evidenceId":"mimo-v26-flash:model-card:terminal-bench-4.0","confidence":"high","higherIsBetter":true,"date":"2026-09-22T00:00:00.000Z","observedAt":"2026-09-22T03:59:47.024Z"},{"source":"xiaomi-mimo","name":"Terminal-Bench 2.1","score":87.6,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal Bench 2.1","sourceUrl":"https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Pro-RL","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"MiMo-V2.6 Flash","configuration":{"protocol":"xiaomi-mimo-v2.6-model-card-2026-09-22","metric":"model-card reported score"},"benchmarkVersion":"2.1","artifactId":"xiaomi-mimo-v26-hf-7e3928885f5658a7","evidenceId":"mimo-v26-flash:model-card:terminal-bench-2.1","confidence":"high","higherIsBetter":true,"date":"2026-09-22T00:00:00.000Z","observedAt":"2026-09-22T03:59:47.024Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3.2","score":37.8843590141754,"canonicalName":"Artificial Analysis Intelligence Index v4.3.2","rawName":"Artificial Analysis Intelligence Index v4.3.2","sourceUrl":"https://artificialanalysis.ai/models/mimo-v2-6-flash","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3.2","modelVariant":"MiMo-V2.6-Flash","exactModelVariant":"MiMo-V2.6-Flash","configuration":{"protocol":"aa-independent-2026-09-v4.3.2","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","components":["AA-Briefcase v1.1","GDPval-AA v2.1","AutomationBench-AA","Terminal-Bench 4.0","SciCode","Humanity's Last Exam","GDP.pdf","CritPt","AA-Omniscience","AA-LCR v1.1"],"modelSettings":{"evaluatorModelId":"127de816-57d3-4037-8484-480a613cd413","evaluatorModelSlug":"mimo-v2-6-flash","reasoning":true,"selection":"Exact non-estimated evaluator row","weightsSourceUrl":"https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL"}},"benchmarkVersion":"4.3.2","artifactId":"aa-mimo-flash-050749c37d0d9452","evidenceId":"mimo-v26-flash:aa:intelligenceIndex:v4.3.2","confidence":"high","higherIsBetter":true,"comparisonEligible":false,"exclusionReason":"BenchmarkProtocolReviewRequired: v4.3.2 is not the globally selected exact protocol","estimated":false,"date":null,"observedAt":"2026-09-30T01:45:14.354Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6abf05d7dbc6c890860c2ff4","name":"Grok 4.7","slug":"grok-4-7","apiIdentifier":"grok-4.7","family":"Grok","version":"4.7","description":"Grok 4.7 is SpaceXAI's proprietary coding and knowledge-work model, available as grok-4.7 on the xAI API. It accepts text and images, returns text, and offers a 500,000-token context window. SpaceXAI's release notes specify no text output limit. Standard global API pricing starts at $2 input and $6 output per million tokens, with separate cached-input and long-context rates.","organizationSlug":"xai","organizationName":"xAI","contextWindow":500000,"maxOutputTokens":null,"inputPricePerMillion":2,"outputPricePerMillion":6,"capabilities":["coding","reasoning","vision","tool-use","structured-output","long-context","context-caching"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-09-21T00:00:00.000Z","tags":["coding","reasoning","vision","long-context","api","grok"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3.2","score":42.2166560223448,"canonicalName":"Artificial Analysis Intelligence Index v4.3.2","rawName":"Artificial Analysis Intelligence Index v4.3.2","sourceUrl":"https://artificialanalysis.ai/models/grok-4-7-low","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3.2","modelVariant":"Grok 4.7 (Low)","exactModelVariant":"Grok 4.7 (Low)","configuration":{"protocol":"aa-independent-2026-09-v4.3.2","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","components":["AA-Briefcase v1.1","GDPval-AA v2.1","AutomationBench-AA","Terminal-Bench 4.0","SciCode","Humanity's Last Exam","GDP.pdf","CritPt","AA-Omniscience","AA-LCR v1.1"],"modelSettings":{"evaluatorModelId":"208c16f0-8c5b-4524-ac8a-ac74c02e80a9","evaluatorModelSlug":"grok-4-7-low","reasoning":true,"effort":"low","selection":"Exact non-estimated evaluator row; one API model, not a separate low-effort product"}},"benchmarkVersion":"4.3.2","artifactId":"aa-grok47-low-3fbefcac9a46ca19","evidenceId":"grok-4-7:aa:intelligenceIndex:v4.3.2:low","confidence":"high","higherIsBetter":true,"comparisonEligible":true,"estimated":false,"date":null,"observedAt":"2026-10-02T01:06:27.411Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a97a7196eda2882a871b21d","name":"MiniMind","slug":"minimind","apiIdentifier":"jingyaogong/minimind","family":"MiniMind","version":"64M","description":"MiniMind is a small open-source language-model training project that demonstrates how to train a 64M-parameter LLM from scratch. The repository is aimed at builders who want a compact, inspectable model-training path rather than a hosted frontier API.","organizationSlug":"jingyaogong","organizationName":"Jingyao Gong","contextWindow":1024,"maxOutputTokens":1024,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","education","training"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-09-02T04:33:26.357Z","tags":["small-llm","training","open-source","education","pytorch"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6aa5f420f39a9367cbe36c80","name":"Claude Fable 5.1","slug":"claude-fable-5-1","apiIdentifier":"claude-fable-5-1","family":"Claude Fable","version":"5.1","description":"Claude Fable 5.1 is Anthropic's model for demanding reasoning and long-horizon agentic work. It supports text and image input, text output, a 1M-token context window and always-on adaptive thinking. Anthropic recommends Claude Opus 5 for most workloads and Fable 5.1 when higher-effort Opus evaluations fall short.","organizationSlug":"anthropic","organizationName":"Anthropic","contextWindow":1000000,"maxOutputTokens":128000,"inputPricePerMillion":10,"outputPricePerMillion":50,"capabilities":["text","vision","reasoning","coding","tool-use","agents","long-context","adaptive-thinking"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-09-01T00:00:00.000Z","tags":["reasoning","coding","agentic","multimodal","long-context","adaptive-thinking"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":53.3737509623252,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"3e87c73e-a257-495e-9730-367a66229811","evaluatorModelSlug":"claude-fable-5-1","reasoning":true,"effort":"max","fallback":"default","selection":"Exact named evaluator configuration, max effort with default fallback","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-cohort","evidenceId":"claude-fable-5-1:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-13T00:29:05.183Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":93.7373737373737,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"3e87c73e-a257-495e-9730-367a66229811","evaluatorModelSlug":"claude-fable-5-1","reasoning":true,"effort":"max","fallback":"default","selection":"Exact named evaluator configuration, max effort with default fallback","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-cohort","evidenceId":"claude-fable-5-1:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-13T00:29:05.183Z"},{"source":"artificial-analysis","name":"SciCode","score":63.078703703703695,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"3e87c73e-a257-495e-9730-367a66229811","evaluatorModelSlug":"claude-fable-5-1","reasoning":true,"effort":"max","fallback":"default","selection":"Exact named evaluator configuration, max effort with default fallback","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-cohort","evidenceId":"claude-fable-5-1:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-13T00:29:05.183Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":91.3857677902622,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"3e87c73e-a257-495e-9730-367a66229811","evaluatorModelSlug":"claude-fable-5-1","reasoning":true,"effort":"max","fallback":"default","selection":"Exact named evaluator configuration, max effort with default fallback","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-cohort","evidenceId":"claude-fable-5-1:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-13T00:29:05.183Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":53.3737509623252},{"_id":"6a9112d0cffc29cb6847741e","name":"Gemini Omni 1.1 Flash","slug":"gemini-omni-1-1-flash","apiIdentifier":"gemini-omni-1.1-flash","family":"Gemini Omni","version":"1.1 Flash","description":"Gemini Omni 1.1 Flash is Google’s high-performance video generation and editing model for the Gemini API. It turns text, images, and short video references into controllable video output with scene extension, interpolation, conversational editing, and upscaling options.","organizationSlug":"google","organizationName":"Google","contextWindow":1048576,"maxOutputTokens":57920,"inputPricePerMillion":1.5,"outputPricePerMillion":17.5,"capabilities":["video-generation","video-editing","text-to-video","image-to-video","video-extension","multimodal-input","cinematic-control"],"category":"Video Model","current":true,"deprecated":false,"releaseDate":"2026-08-27T00:00:00.000Z","tags":["video-generation","gemini","multimodal","video-editing","text-to-video","image-to-video"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a9112d0cffc29cb6847741f","name":"GLM-5.3-Flash","slug":"glm-5-3-flash","apiIdentifier":"z-ai/glm-5.3-flash","family":"GLM","version":"5.3 Flash","description":"GLM-5.3-Flash is Z.ai’s open-weight GLM-5 model for efficient coding, long-horizon agent tasks, and long-context multimodal reasoning. It uses a hybrid sparse and linear attention design with 320B total parameters and 18B active parameters.","organizationSlug":"z-ai","organizationName":"Z.ai","contextWindow":1310720,"maxOutputTokens":131072,"inputPricePerMillion":0.075,"outputPricePerMillion":0.25,"capabilities":["text","vision","video-input","code","tool-use","reasoning","long-context","open-weights"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-08-26T00:00:00.000Z","tags":["open-weights","reasoning","coding","multimodal","long-context","glm"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1.1","score":57,"date":"2026-08-26T00:00:00.000Z"},{"source":"official","name":"MMLU","score":88.1,"date":"2026-08-26T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":41.9074,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GLM-5.3-Flash","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"19496b81-9f41-4214-a77a-1df803b3c5ae","evaluatorModelSlug":"glm-5-3-flash","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"glm-5-3-flash:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":91.2121,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5.3-Flash","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"19496b81-9f41-4214-a77a-1df803b3c5ae","evaluatorModelSlug":"glm-5-3-flash","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"glm-5-3-flash:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":51.6204,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5.3-Flash","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"19496b81-9f41-4214-a77a-1df803b3c5ae","evaluatorModelSlug":"glm-5-3-flash","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"glm-5-3-flash:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":84.2697,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5.3-Flash","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"19496b81-9f41-4214-a77a-1df803b3c5ae","evaluatorModelSlug":"glm-5-3-flash","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"glm-5-3-flash:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":57},{"_id":"6a9573ef366bb45f1153882c","name":"Apodex 1.1","slug":"apodex-1-1","apiIdentifier":"apodex-1.1","family":"Apodex","version":"1.1","description":"Apodex 1.1 is a proprietary long-context reasoning model for complex professional work, research, coding, and evidence-heavy analysis through a first-party API.","organizationSlug":"apodex","organizationName":"Apodex","contextWindow":262144,"maxOutputTokens":262144,"inputPricePerMillion":0.24,"outputPricePerMillion":2.4,"capabilities":["text","reasoning","code","long-context","streaming"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-08-24T00:00:00.000Z","tags":["reasoning","long-context","api","proprietary","agentic"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1.1","score":44,"rawName":"Artificial Analysis Intelligence Index","sourceUrl":"https://artificialanalysis.ai/models/apodex-1-1","sourceType":"independent-benchmark","unit":"index-points","scale":"Artificial Analysis Intelligence Index v4.1.1","modelVariant":"Apodex 1.1","artifactId":"aa-apodex-11.txt","evidenceId":"src_aa_apodex","confidence":"high","date":"2026-08-30T00:00:00.000Z","observedDate":"2026-08-31T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":86.3636,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Apodex 1.1","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"08242385-ee91-4de6-be7f-271a6d9780ce","evaluatorModelSlug":"apodex-1-1","reasoning":true,"effort":null,"selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"apodex-1-1:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":45.4861,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Apodex 1.1","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"08242385-ee91-4de6-be7f-271a6d9780ce","evaluatorModelSlug":"apodex-1-1","reasoning":true,"effort":null,"selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"apodex-1-1:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":69.6629,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Apodex 1.1","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"08242385-ee91-4de6-be7f-271a6d9780ce","evaluatorModelSlug":"apodex-1-1","reasoning":true,"effort":null,"selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"apodex-1-1:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":44},{"_id":"6a732c0d3bc1ce213a0a3338","name":"DeepSeek V4 Pro","slug":"deepseek-v4-pro","apiIdentifier":"deepseek-v4-pro","family":"DeepSeek V4","version":"Pro 0813","description":"DeepSeek V4 Pro 0813 is an open-weight Mixture-of-Experts reasoning model for long-context coding, tool use, and agentic workloads, with native Responses API support through the DeepSeek API.","organizationSlug":"deepseek","organizationName":"DeepSeek","contextWindow":1048576,"maxOutputTokens":393216,"inputPricePerMillion":1.32,"outputPricePerMillion":3.96,"capabilities":["reasoning","coding","tool-use","structured-output","long-context","agentic","responses-api"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-08-13T00:00:00.000Z","tags":["reasoning","open-weights","coding","long-context","tool-use","agentic","responses-api"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1","score":44,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1","sourceUrl":"https://artificialanalysis.ai/models/deepseek-v4-pro/","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"DeepSeek V4 Pro (Reasoning, Max Effort)","evidenceId":"sb-aa-deepseek-v4-pro-model","confidence":"high","date":"2026-08-05T00:00:00.000Z","observedDate":"2026-08-05T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1.1","score":53,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1.1","sourceUrl":"https://artificialanalysis.ai/models/deepseek-v4-pro","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"DeepSeek V4 Pro 0813 (Reasoning, Max Effort)","evidenceId":"sb-aa-deepseek-v4-pro","confidence":"high","date":"2026-08-13T00:00:00.000Z","observedDate":"2026-08-13T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":36.2829,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"DeepSeek V4 Pro 0813 (Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"83173329-c09a-41f1-a028-a282a5f908d7","evaluatorModelSlug":"deepseek-v4-pro","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"deepseek-v4-pro:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":92.8283,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V4 Pro 0813 (Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"83173329-c09a-41f1-a028-a282a5f908d7","evaluatorModelSlug":"deepseek-v4-pro","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"deepseek-v4-pro:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":51.0417,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V4 Pro 0813 (Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"83173329-c09a-41f1-a028-a282a5f908d7","evaluatorModelSlug":"deepseek-v4-pro","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"deepseek-v4-pro:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":78.6517,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V4 Pro 0813 (Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"83173329-c09a-41f1-a028-a282a5f908d7","evaluatorModelSlug":"deepseek-v4-pro","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"deepseek-v4-pro:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3.2","score":20.3605101726847,"canonicalName":"Artificial Analysis Intelligence Index v4.3.2","rawName":"Artificial Analysis Intelligence Index v4.3.2","sourceUrl":"https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3.2","modelVariant":"DeepSeek V4 Pro 0813 (Non-reasoning)","exactModelVariant":"DeepSeek V4 Pro 0813 (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3.2","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","components":["AA-Briefcase v1.1","GDPval-AA v2.1","AutomationBench-AA","Terminal-Bench 4.0","SciCode","Humanity's Last Exam","GDP.pdf","CritPt","AA-Omniscience","AA-LCR v1.1"],"modelSettings":{"evaluatorModelId":"d4f5b250-be37-4801-85e6-32fda4179337","evaluatorModelSlug":"deepseek-v4-pro-non-reasoning","reasoning":false,"effort":"none","selection":"Exact non-estimated evaluator row; same official deepseek-v4-pro API model with thinking disabled","weightsSourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813"}},"benchmarkVersion":"4.3.2","artifactId":"aa-deepseek-nonreasoning-1abd164f05f8b521","evidenceId":"deepseek-v4-pro:aa:nonreasoning:v4.3.2","confidence":"high","higherIsBetter":true,"comparisonEligible":false,"exclusionReason":"BenchmarkProtocolReviewRequired: v4.3.2 is not the globally selected exact protocol","estimated":false,"date":null,"observedAt":"2026-09-30T13:13:58.224Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":53},{"_id":"6a7f0dc8e9c3505fe5812230","name":"Gemini 3.7 Flash","slug":"gemini-37-flash","apiIdentifier":"gemini-3.7-flash","family":"Gemini","version":"3.7 Flash","description":"Gemini 3.7 Flash is Google's multimodal reasoning model for coding, agentic workflows, and complex knowledge work. It supports text, image, audio, and video input, customizable thinking, a 1M context window, 64K text output, and access through the Gemini API.","organizationSlug":"google","organizationName":"Google","contextWindow":1048576,"maxOutputTokens":65536,"inputPricePerMillion":0.75,"outputPricePerMillion":3.75,"capabilities":["reasoning","coding","agentic","tool-use","multimodal","long-context","image-understanding","audio-understanding","video-understanding"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-08-13T00:00:00.000Z","tags":["reasoning","coding","agentic","multimodal","long-context","gemini-api"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1.1","score":56,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1.1","sourceUrl":"https://artificialanalysis.ai/models/gemini-3-7-flash","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"Gemini 3.7 Flash (high)","evidenceId":"sb-aa-gemini-3-7-model","confidence":"high","date":"2026-08-13T00:00:00.000Z","observedDate":"2026-08-14T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/gemini-3-7-flash","name":"Artificial Analysis Intelligence Index v4.3","score":39,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/gemini-3-7-flash","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Gemini 3.7 Flash (high)","configuration":{"protocol":"Artificial Analysis Intelligence Index v4.3; exact effort/fallback in model variant"},"benchmarkVersion":"4.3","artifactId":"e1811d3243678561","evidenceId":"gemini-37-flash:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:28:41.597Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":39.4295,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Gemini 3.7 Flash (high)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b2331108-72ed-415a-82d1-188633875bbc","evaluatorModelSlug":"gemini-3-7-flash","reasoning":true,"effort":"high","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gemini-37-flash:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":94.5455,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.7 Flash (high)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b2331108-72ed-415a-82d1-188633875bbc","evaluatorModelSlug":"gemini-3-7-flash","reasoning":true,"effort":"high","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemini-37-flash:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":57.1759,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.7 Flash (high)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b2331108-72ed-415a-82d1-188633875bbc","evaluatorModelSlug":"gemini-3-7-flash","reasoning":true,"effort":"high","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemini-37-flash:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":85.7678,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.7 Flash (high)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b2331108-72ed-415a-82d1-188633875bbc","evaluatorModelSlug":"gemini-3-7-flash","reasoning":true,"effort":"high","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gemini-37-flash:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":56},{"_id":"6a7e9d142e4bf3014a3284fd","name":"Cactus Needle 2","slug":"cactus-needle-2","apiIdentifier":"Cactus-Compute/needle2","family":"Needle","version":"2","description":"Cactus Needle 2 is a 45M-parameter, 14 MB agentic model for tiny devices. It focuses on tool calling, device control, schema-constrained JSON, and structured extraction; Cactus reports 28 MB session RAM, 800+ tok/s prefill, and 500+ tok/s decode on Raspberry Pi 5. Its 256-token sliding attention window is not a total context or output limit; those ceilings are not established by the model documentation.","organizationSlug":"cactus-compute","organizationName":"Cactus Compute","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["tool-use","structured-output","json-schema","device-control","edge-inference","offline"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-08-11T00:00:00.000Z","tags":["edge-llm","tool-calling","structured-output","tiny-model","on-device-ai","apache-2"],"benchmarks":[{"source":"cactus-compute","name":"Mobile Actions Ordered Strict Exact Match","score":63.7,"canonicalName":"Mobile Actions","rawName":"Mobile Actions (961 rows) Accuracy","sourceUrl":"https://cactuscompute.com/needle","sourceType":"official-benchmark","unit":"percent","scale":"0-100 higher is better","modelVariant":"Needle 2 (CQ2-bit, shipped C++ engine, tool retrieval on)","evidenceId":"sb-cactus-needle","confidence":"high","date":null,"observedDate":"2026-08-14T00:00:00.000Z"},{"source":"cactus-compute","name":"BFCL v4 Single-Turn Overall","score":42.6,"canonicalName":"Berkeley Function Calling Leaderboard v4","rawName":"BFCL v4 single-turn Overall","sourceUrl":"https://cactuscompute.com/needle","sourceType":"official-benchmark","unit":"percent","scale":"0-100 higher is better","modelVariant":"Needle 2 (CQ2-bit, shipped C++ engine)","evidenceId":"sb-cactus-needle","confidence":"high","date":null,"observedDate":"2026-08-14T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a7c66530a00be26a915c14e","name":"Nemotron 3.5 Lightning 30B A3B","slug":"nemotron-35-lightning-30b-a3b","apiIdentifier":"nvidia/nemotron-3.5-lightning-30b-a3b","family":"Nemotron","version":"3.5 Lightning 30B A3B","description":"Nemotron 3.5 Lightning is NVIDIA open-weight 30B-A3B reasoning model for fast, long-running agents. Its hybrid Mamba-2, mixture-of-experts, and attention architecture supports function calling, coding, tool use, long context, and efficient local or serverless deployment.","organizationSlug":"nvidia","organizationName":"NVIDIA","contextWindow":1048576,"maxOutputTokens":32768,"inputPricePerMillion":0.1,"outputPricePerMillion":0.95,"capabilities":["text","reasoning","coding","tool-use","function-calling","long-context","agentic","local-deployment","open-weights"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-08-11T00:00:00.000Z","tags":["reasoning","agentic","long-context","open-weight","function-calling","moe","nvfp4"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1.1","score":24,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1.1","sourceUrl":"https://artificialanalysis.ai/models/nemotron-3-5-lightning","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"Nemotron 3.5 Lightning","evidenceId":"sb-aa-nemotron-3-5-lightning-model","confidence":"high","date":"2026-08-11T00:00:00.000Z","observedDate":"2026-08-12T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/nemotron-3-5-lightning","name":"Artificial Analysis Intelligence Index v4.3","score":14,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/nemotron-3-5-lightning","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Nemotron 3.5 Lightning","configuration":{"protocol":"Artificial Analysis Intelligence Index v4.3; exact effort/fallback in model variant"},"benchmarkVersion":"4.3","artifactId":"1aedccf961db94d2","evidenceId":"nemotron-35-lightning-30b-a3b:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:28:38.285Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":13.6353,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Nemotron 3.5 Lightning","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"29976311-665a-4b2f-ac72-557c33e0758e","evaluatorModelSlug":"nemotron-3-5-lightning","reasoning":true,"effort":null,"selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"nemotron-35-lightning-30b-a3b:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":74.3434,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Nemotron 3.5 Lightning","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"29976311-665a-4b2f-ac72-557c33e0758e","evaluatorModelSlug":"nemotron-3-5-lightning","reasoning":true,"effort":null,"selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"nemotron-35-lightning-30b-a3b:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":32.0602,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Nemotron 3.5 Lightning","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"29976311-665a-4b2f-ac72-557c33e0758e","evaluatorModelSlug":"nemotron-3-5-lightning","reasoning":true,"effort":null,"selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"nemotron-35-lightning-30b-a3b:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":24.3446,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Nemotron 3.5 Lightning","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"29976311-665a-4b2f-ac72-557c33e0758e","evaluatorModelSlug":"nemotron-3-5-lightning","reasoning":true,"effort":null,"selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"nemotron-35-lightning-30b-a3b:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":24},{"_id":"6a7e9d142e4bf3014a3284fc","name":"IndexTTS","slug":"indextts","apiIdentifier":"IndexTeam/IndexTTS-2.5","family":"IndexTTS","version":"2.5","description":"IndexTTS is a controllable zero-shot text-to-speech model family from IndexTeam and Bilibili. The 2.5 release clones voices from a single reference clip, supports Chinese, English, Japanese, Spanish, and Arabic, and reports a 2.28x real-time-factor improvement over IndexTTS 2 in its technical report.","organizationSlug":"indexteam","organizationName":"IndexTeam","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text-to-speech","voice-cloning","multilingual","emotion-control","pronunciation-control","speed-control"],"category":"Audio Model","current":true,"deprecated":false,"releaseDate":"2026-08-10T00:00:00.000Z","tags":["text-to-speech","voice-cloning","speech-synthesis","multilingual","emotion-control","open-weights"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a76b2666ce9e009bf93b988","name":"Maple-Preview","slug":"maple-preview","apiIdentifier":"deepgrove/maple-preview","family":"Maple","version":"Preview","description":"Maple-Preview is DeepGrove’s open-source 20B-A1B ternary-weight reasoning LLM for efficient on-device inference. The model card reports 20.2B total parameters, 1.49B active parameters, a 131,072-token context window, and 5.31 GB checkpoint size. The model card specifies 131,072 tokens of context but does not establish a separate maximum output-token limit.","organizationSlug":"deepgrove","organizationName":"DeepGrove","contextWindow":131072,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","reasoning","code","on-device-inference"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-08-08T04:36:54.568Z","tags":["llm","reasoning-model","on-device-ai","ternary-weights","low-precision","moe"],"benchmarks":[{"source":"DeepGrove model card","name":"LiveCodeBench v6","score":75.1,"date":"2026-08-08T00:00:00.000Z"},{"source":"DeepGrove model card","name":"AIME 2026","score":87.5,"date":"2026-08-08T00:00:00.000Z"},{"source":"DeepGrove model card","name":"HMMT 2026","score":78.8,"date":"2026-08-08T00:00:00.000Z"},{"source":"DeepGrove model card","name":"GPQA Diamond","score":73.5,"date":"2026-08-08T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a689fdfeb09e1180791b5ed","name":"Claude Opus 5","slug":"claude-opus-5","apiIdentifier":"claude-opus-5","family":"Claude Opus","version":"5","description":"Claude Opus 5 is Anthropic's Opus-class model for complex agentic coding and enterprise work. Official Claude docs list the claude-opus-5 API ID, 1M token context window, 128k max output, adaptive thinking, and $5 input / $25 output pricing per million tokens.","organizationSlug":"anthropic","organizationName":"Anthropic","contextWindow":1000000,"maxOutputTokens":128000,"inputPricePerMillion":5,"outputPricePerMillion":25,"capabilities":["text","vision","reasoning","coding","tool-use","agents","long-context"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-07-24T00:00:00.000Z","tags":["reasoning","coding","agentic","multimodal","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index","score":61,"rawName":"Claude Opus 5 (Adaptive Reasoning, Max Effort)","sourceUrl":"https://artificialanalysis.ai/leaderboards/models","sourceType":"benchmark-leaderboard","unit":"index","scale":"Artificial Analysis Intelligence Index","exactModelVariant":"Claude Opus 5 (Adaptive Reasoning, Max Effort)","evidenceId":"2026-07-28-artificial-analysis-leaderboard","confidence":"high","date":"2026-07-28T00:00:00.000Z","observedDate":"2026-07-28T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1.1","score":63,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"Claude Opus 5 (Adaptive Reasoning, Max Effort)","exactModelVariant":"Claude Opus 5 (Adaptive Reasoning, Max Effort)","evidenceId":"sb-aa-claude-opus-5","confidence":"high","date":"2026-08-15T00:00:00.000Z","observedDate":"2026-08-15T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/claude-opus-5","name":"Artificial Analysis Intelligence Index v4.3","score":51,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Claude Opus 5 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"Artificial Analysis Intelligence Index v4.3; exact effort/fallback in model variant"},"benchmarkVersion":"4.3","artifactId":"863db5441ba7ca03","evidenceId":"claude-opus-5:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:28:34.246Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":50.7002,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Claude Opus 5 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b8fc61f7-5e9a-49e6-8547-6ac56db24627","evaluatorModelSlug":"claude-opus-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"claude-opus-5:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":93.2323,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Opus 5 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b8fc61f7-5e9a-49e6-8547-6ac56db24627","evaluatorModelSlug":"claude-opus-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"claude-opus-5:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":56.3657,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Opus 5 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b8fc61f7-5e9a-49e6-8547-6ac56db24627","evaluatorModelSlug":"claude-opus-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"claude-opus-5:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":89.1386,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Opus 5 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b8fc61f7-5e9a-49e6-8547-6ac56db24627","evaluatorModelSlug":"claude-opus-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"claude-opus-5:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":63},{"_id":"6a674e6eef8be0792f0212d0","name":"Gemini 3.6 Flash","slug":"gemini-36-flash","apiIdentifier":"gemini-3.6-flash","family":"Gemini","version":"3.6 Flash","description":"Gemini 3.6 Flash is Google DeepMind general availability Flash model for coding, agentic execution, spatial reasoning, multimodal understanding, and long-context production workflows. Official DeepMind evidence lists 1M input tokens, 64k output tokens, text, image, video, audio, and PDF inputs, text output, tool use, Gemini API availability, and $1.50 input / $7.50 output pricing per million tokens.","organizationSlug":"google","organizationName":"Google","contextWindow":1000000,"maxOutputTokens":64000,"inputPricePerMillion":1.5,"outputPricePerMillion":7.5,"capabilities":["text","vision","audio-input","video-input","pdf-input","reasoning","coding","tool-use","search-grounding","computer-use","structured-output"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-07-21T00:00:00.000Z","tags":["gemini","multimodal","reasoning","agentic-coding","long-context","google-ai"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index","score":50,"rawName":"Artificial Analysis Intelligence Index","sourceUrl":"https://artificialanalysis.ai/models/gemini-3-6-flash","sourceType":"benchmark-leaderboard","unit":"index points","scale":"0-100","modelVariant":"Gemini 3.6 Flash","evidenceId":"aa-gemini-36-flash","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-07-27T00:00:00.000Z"},{"source":"google-deepmind","name":"SWE-Bench Pro","score":58.7,"rawName":"SWE-Bench Pro (Public) Diverse agentic coding tasks","sourceUrl":"https://deepmind.google/models/gemini/flash/","sourceType":"official-model-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.6 Flash","evidenceId":"deepmind-gemini-36-flash","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-07-27T00:00:00.000Z"},{"source":"google-deepmind","name":"Terminal-Bench 2.1","score":78,"rawName":"Terminal-bench 2.1 Agentic terminal coding; Terminus-2 harness","sourceUrl":"https://deepmind.google/models/gemini/flash/","sourceType":"official-model-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.6 Flash","evidenceId":"deepmind-gemini-36-flash","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-07-27T00:00:00.000Z"},{"source":"google-deepmind","name":"GDPVal-AA v2","score":1421,"rawName":"GDPVal-AA v2 Knowledge work","sourceUrl":"https://deepmind.google/models/gemini/flash/","sourceType":"official-model-benchmark","unit":"Elo","scale":"Elo","modelVariant":"Gemini 3.6 Flash","evidenceId":"deepmind-gemini-36-flash","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-07-27T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/gemini-3-6-flash","name":"Artificial Analysis Intelligence Index v4.3","score":34,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/gemini-3-6-flash","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Gemini 3.6 Flash (high)","configuration":{"protocol":"Artificial Analysis Intelligence Index v4.3; exact effort/fallback in model variant"},"benchmarkVersion":"4.3","artifactId":"cb404c6598b9c132","evidenceId":"gemini-36-flash:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:28:30.999Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":34.3396,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Gemini 3.6 Flash (high)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"78332c88-fe60-42c8-af9d-3617e44cf1f5","evaluatorModelSlug":"gemini-3-6-flash","reasoning":true,"effort":"high","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gemini-36-flash:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":92.8283,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.6 Flash (high)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"78332c88-fe60-42c8-af9d-3617e44cf1f5","evaluatorModelSlug":"gemini-3-6-flash","reasoning":true,"effort":"high","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemini-36-flash:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":53.3565,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.6 Flash (high)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"78332c88-fe60-42c8-af9d-3617e44cf1f5","evaluatorModelSlug":"gemini-3-6-flash","reasoning":true,"effort":"high","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemini-36-flash:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":77.5281,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.6 Flash (high)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"78332c88-fe60-42c8-af9d-3617e44cf1f5","evaluatorModelSlug":"gemini-3-6-flash","reasoning":true,"effort":"high","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gemini-36-flash:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":50},{"_id":"6a674e71ef8be0792f0212d1","name":"Gemini 3.5 Flash-Lite","slug":"gemini-35-flash-lite","apiIdentifier":"gemini-3.5-flash-lite","family":"Gemini","version":"3.5 Flash-Lite","description":"Gemini 3.5 Flash-Lite is Google DeepMind general availability low-latency model for high-volume agentic systems, coding, document parsing, translation, and structured automation. Official DeepMind evidence lists 1M input tokens, 64k output tokens, text, image, video, audio, and PDF inputs, text output, tool use, Gemini API availability, and $0.30 input / $2.50 output pricing per million tokens.","organizationSlug":"google","organizationName":"Google","contextWindow":1000000,"maxOutputTokens":64000,"inputPricePerMillion":0.3,"outputPricePerMillion":2.5,"capabilities":["text","vision","audio-input","video-input","pdf-input","reasoning","coding","tool-use","search-grounding","computer-use","structured-output"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-07-21T00:00:00.000Z","tags":["gemini","flash-lite","multimodal","low-latency","agentic-workflows","google-ai"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index","score":36,"rawName":"Artificial Analysis Intelligence Index","sourceUrl":"https://artificialanalysis.ai/articles/gemini-3-6-flash-3-5-flash-lite-halving-time","sourceType":"benchmark-leaderboard","unit":"index points","scale":"0-100","modelVariant":"Gemini 3.5 Flash-Lite","evidenceId":"aa-gemini-flash-lite-article","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-07-27T00:00:00.000Z"},{"source":"google-deepmind","name":"SWE-Bench Pro","score":54.2,"rawName":"SWE-Bench Pro (Public) Diverse agentic coding tasks","sourceUrl":"https://deepmind.google/models/gemini/flash-lite/","sourceType":"official-model-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.5 Flash-Lite","evidenceId":"deepmind-gemini-35-flash-lite","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-07-27T00:00:00.000Z"},{"source":"google-deepmind","name":"Terminal-Bench 2.1","score":54,"rawName":"Terminal-bench 2.1 Agentic terminal coding; Terminus-2 harness","sourceUrl":"https://deepmind.google/models/gemini/flash-lite/","sourceType":"official-model-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.5 Flash-Lite","evidenceId":"deepmind-gemini-35-flash-lite","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-07-27T00:00:00.000Z"},{"source":"google-deepmind","name":"GDPVal-AA v2","score":1140,"rawName":"GDPVal-AA v2 Knowledge work","sourceUrl":"https://deepmind.google/models/gemini/flash-lite/","sourceType":"official-model-benchmark","unit":"Elo","scale":"Elo","modelVariant":"Gemini 3.5 Flash-Lite","evidenceId":"deepmind-gemini-35-flash-lite","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-07-27T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":22.6607,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Gemini 3.5 Flash-Lite","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"093b9df2-ebee-4309-943f-5a7e88360adf","evaluatorModelSlug":"gemini-3-5-flash-lite","reasoning":true,"effort":"high","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gemini-35-flash-lite:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":83.8384,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.5 Flash-Lite","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"093b9df2-ebee-4309-943f-5a7e88360adf","evaluatorModelSlug":"gemini-3-5-flash-lite","reasoning":true,"effort":"high","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemini-35-flash-lite:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":41.3194,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.5 Flash-Lite","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"093b9df2-ebee-4309-943f-5a7e88360adf","evaluatorModelSlug":"gemini-3-5-flash-lite","reasoning":true,"effort":"high","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemini-35-flash-lite:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":53.5581,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemini 3.5 Flash-Lite","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"093b9df2-ebee-4309-943f-5a7e88360adf","evaluatorModelSlug":"gemini-3-5-flash-lite","reasoning":true,"effort":"high","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gemini-35-flash-lite:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":36},{"_id":"6a635ae619b4c6fbd0b1a09c","name":"GPT-5.6 Sol","slug":"gpt-56-sol","apiIdentifier":"gpt-5.6-sol","family":"GPT-5.6","version":"Sol","description":"GPT-5.6 Sol is OpenAI frontier GPT-5.6 model for complex professional work across coding, knowledge work, science, cybersecurity, and agentic workflows. OpenAI model docs list it as the GPT-5.6 alias target with function, web search, file search, and computer use tools, 1.05M context, and 128K max output.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":4,"outputPricePerMillion":20,"capabilities":["text","vision","reasoning","coding","tool-use","web-search","file-search","computer-use","agents"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-07-09T00:00:00.000Z","tags":["reasoning","coding","agentic","multimodal","tool-use"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index","score":59,"rawName":"Artificial Analysis Intelligence Index","sourceUrl":"https://artificialanalysis.ai/articles/gpt-5-6-has-landed","sourceType":"benchmark-leaderboard","unit":"index","scale":"0-100","evidenceId":"aa-gpt-5-6-article","confidence":"high","date":"2026-07-09T00:00:00.000Z","observedDate":"2026-07-24T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Coding Agent Index","score":80,"rawName":"Artificial Analysis Coding Agent Index","sourceUrl":"https://artificialanalysis.ai/articles/gpt-5-6-has-landed","sourceType":"benchmark-leaderboard","unit":"index","scale":"0-100","evidenceId":"aa-gpt-5-6-article","confidence":"high","date":"2026-07-09T00:00:00.000Z","observedDate":"2026-07-24T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1.1","score":61,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1.1","sourceUrl":"https://artificialanalysis.ai/models/gpt-5-6-sol","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"GPT-5.6 Sol (max)","exactModelVariant":"GPT-5.6 Sol (max)","evidenceId":"sb-aa-gpt-5-6-sol","confidence":"high","date":"2026-08-15T00:00:00.000Z","observedDate":"2026-08-15T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/gpt-5-6-sol","name":"Artificial Analysis Intelligence Index v4.3","score":47,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/gpt-5-6-sol","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GPT-5.6 Sol (max)","configuration":{"protocol":"Artificial Analysis Intelligence Index v4.3; exact effort/fallback in model variant"},"benchmarkVersion":"4.3","artifactId":"29d2fef1679c11c2","evidenceId":"gpt-56-sol:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:28:29.691Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":47.0614,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GPT-5.6 Sol (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d93edfe8-bf35-49ad-b56e-b18116142a1c","evaluatorModelSlug":"gpt-5-6-sol","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gpt-56-sol:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":94.1414,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.6 Sol (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d93edfe8-bf35-49ad-b56e-b18116142a1c","evaluatorModelSlug":"gpt-5-6-sol","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-56-sol:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":57.0602,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.6 Sol (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d93edfe8-bf35-49ad-b56e-b18116142a1c","evaluatorModelSlug":"gpt-5-6-sol","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-56-sol:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":88.015,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.6 Sol (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d93edfe8-bf35-49ad-b56e-b18116142a1c","evaluatorModelSlug":"gpt-5-6-sol","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gpt-56-sol:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":61},{"_id":"6a635ae819b4c6fbd0b1a09d","name":"GPT-5.6 Terra","slug":"gpt-56-terra","apiIdentifier":"gpt-5.6-terra","family":"GPT-5.6","version":"Terra","description":"GPT-5.6 Terra is OpenAI balanced GPT-5.6 model for teams that need strong reasoning and agentic performance at lower cost than Sol. OpenAI model docs list a 1.05M token context window, 128K max output, and support for functions, web search, file search, and computer use.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":2,"outputPricePerMillion":12,"capabilities":["text","vision","reasoning","coding","tool-use","web-search","file-search","computer-use","agents"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-07-09T00:00:00.000Z","tags":["reasoning","coding","agentic","multimodal","cost-efficient"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index","score":55,"rawName":"Artificial Analysis Intelligence Index","sourceUrl":"https://artificialanalysis.ai/articles/gpt-5-6-has-landed","sourceType":"benchmark-leaderboard","unit":"index","scale":"0-100","evidenceId":"aa-gpt-5-6-article","confidence":"high","date":"2026-07-09T00:00:00.000Z","observedDate":"2026-07-24T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Coding Agent Index","score":77.4,"rawName":"Artificial Analysis Coding Agent Index","sourceUrl":"https://artificialanalysis.ai/articles/gpt-5-6-has-landed","sourceType":"benchmark-leaderboard","unit":"index","scale":"0-100","evidenceId":"aa-gpt-5-6-article","confidence":"high","date":"2026-07-09T00:00:00.000Z","observedDate":"2026-07-24T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":42.2515,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GPT-5.6 Terra (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"bcf8db0a-3bb6-4d82-9516-0f57370c85a6","evaluatorModelSlug":"gpt-5-6-terra","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gpt-56-terra:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":92.5253,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.6 Terra (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"bcf8db0a-3bb6-4d82-9516-0f57370c85a6","evaluatorModelSlug":"gpt-5-6-terra","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-56-terra:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":54.9769,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.6 Terra (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"bcf8db0a-3bb6-4d82-9516-0f57370c85a6","evaluatorModelSlug":"gpt-5-6-terra","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-56-terra:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":88.015,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.6 Terra (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"bcf8db0a-3bb6-4d82-9516-0f57370c85a6","evaluatorModelSlug":"gpt-5-6-terra","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gpt-56-terra:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":55},{"_id":"6a635aeb19b4c6fbd0b1a09e","name":"GPT-5.6 Luna","slug":"gpt-56-luna","apiIdentifier":"gpt-5.6-luna","family":"GPT-5.6","version":"Luna","description":"GPT-5.6 Luna is OpenAI lower-cost GPT-5.6 model for high-volume reasoning, coding, and agentic workloads. OpenAI model docs list the same 1.05M token context window, 128K max output, and major tool support as the larger GPT-5.6 variants.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":0.2,"outputPricePerMillion":1.2,"capabilities":["text","vision","reasoning","coding","tool-use","web-search","file-search","computer-use","agents"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-07-09T00:00:00.000Z","tags":["reasoning","coding","agentic","multimodal","low-cost"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index","score":51,"rawName":"Artificial Analysis Intelligence Index","sourceUrl":"https://artificialanalysis.ai/articles/gpt-5-6-has-landed","sourceType":"benchmark-leaderboard","unit":"index","scale":"0-100","evidenceId":"aa-gpt-5-6-article","confidence":"high","date":"2026-07-09T00:00:00.000Z","observedDate":"2026-07-24T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Coding Agent Index","score":74.6,"rawName":"Artificial Analysis Coding Agent Index","sourceUrl":"https://artificialanalysis.ai/articles/gpt-5-6-has-landed","sourceType":"benchmark-leaderboard","unit":"index","scale":"0-100","evidenceId":"aa-gpt-5-6-article","confidence":"high","date":"2026-07-09T00:00:00.000Z","observedDate":"2026-07-24T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":37.5048,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GPT-5.6 Luna (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"426d24c8-49ae-482a-b4a8-20f1c53f21c1","evaluatorModelSlug":"gpt-5-6-luna","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gpt-56-luna:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":91.1111,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.6 Luna (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"426d24c8-49ae-482a-b4a8-20f1c53f21c1","evaluatorModelSlug":"gpt-5-6-luna","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-56-luna:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":53.588,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.6 Luna (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"426d24c8-49ae-482a-b4a8-20f1c53f21c1","evaluatorModelSlug":"gpt-5-6-luna","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-56-luna:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":80.8989,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.6 Luna (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"426d24c8-49ae-482a-b4a8-20f1c53f21c1","evaluatorModelSlug":"gpt-5-6-luna","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gpt-56-luna:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":51},{"_id":"6a723bdcbbf320af3243555e","name":"Muse Spark 1.1","slug":"muse-spark-11","apiIdentifier":"muse-spark-1.1","family":"Muse Spark","version":"1.1","description":"Muse Spark 1.1 is Meta's multimodal reasoning model for agentic coding, computer-use workflows, tool calling, and long-context tasks through the Meta Model API.","organizationSlug":"meta","organizationName":"Meta","contextWindow":1048576,"maxOutputTokens":131072,"inputPricePerMillion":1.25,"outputPricePerMillion":4.25,"capabilities":["reasoning","tool-use","computer-use","coding","multimodal","long-context","structured-output"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-07-09T00:00:00.000Z","tags":["reasoning","agentic-ai","coding","multimodal","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1","score":51,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1","sourceUrl":"https://artificialanalysis.ai/models/muse-spark-1-1/","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"Muse Spark 1.1 (xhigh)","evidenceId":"sb-aa-muse-spark-11-model","confidence":"high","date":"2026-08-04T00:00:00.000Z","observedDate":"2026-08-04T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/muse-spark-1-1/","name":"Artificial Analysis Intelligence Index v4.3","score":34,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/muse-spark-1-1/","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Muse Spark 1.1 (xhigh)","configuration":{"protocol":"Artificial Analysis Intelligence Index v4.3; exact effort/fallback in model variant"},"benchmarkVersion":"4.3","artifactId":"a369345bd139d81d","evidenceId":"muse-spark-11:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:28:35.476Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":34.2658,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Muse Spark 1.1 (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d61493e2-43ff-460c-9054-eaf6ba317360","evaluatorModelSlug":"muse-spark-1-1","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"muse-spark-11:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":89.798,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Muse Spark 1.1 (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d61493e2-43ff-460c-9054-eaf6ba317360","evaluatorModelSlug":"muse-spark-1-1","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"muse-spark-11:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":58.7963,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Muse Spark 1.1 (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d61493e2-43ff-460c-9054-eaf6ba317360","evaluatorModelSlug":"muse-spark-1-1","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"muse-spark-11:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":77.9026,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Muse Spark 1.1 (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d61493e2-43ff-460c-9054-eaf6ba317360","evaluatorModelSlug":"muse-spark-1-1","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"muse-spark-11:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":51},{"_id":"6a635aed19b4c6fbd0b1a09f","name":"Claude Sonnet 5","slug":"claude-sonnet-5","apiIdentifier":"claude-sonnet-5","family":"Claude Sonnet","version":"5","description":"Claude Sonnet 5 is Anthropic next-generation Sonnet model and a drop-in upgrade from Sonnet 4.6. Anthropic docs state that adaptive thinking is on by default, manual extended thinking is removed, sampling parameter overrides return errors, and the model supports a 1M token context window with 128K max output.","organizationSlug":"anthropic","organizationName":"Anthropic","contextWindow":1000000,"maxOutputTokens":128000,"inputPricePerMillion":2,"outputPricePerMillion":10,"capabilities":["text","vision","reasoning","coding","tool-use","extended-thinking","agents"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-06-30T00:00:00.000Z","tags":["reasoning","coding","agentic","multimodal","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index","score":53,"rawName":"Artificial Analysis Intelligence Index","sourceUrl":"https://artificialanalysis.ai/models/claude-sonnet-5","sourceType":"benchmark-leaderboard","unit":"index","scale":"0-100","evidenceId":"aa-claude-sonnet-5-model","confidence":"high","date":"2026-06-30T00:00:00.000Z","observedDate":"2026-07-24T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/claude-sonnet-5","name":"Artificial Analysis Intelligence Index v4.3","score":38,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-sonnet-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Claude Sonnet 5 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"Artificial Analysis Intelligence Index v4.3; exact effort/fallback in model variant"},"benchmarkVersion":"4.3","artifactId":"83acbaa744d20994","evidenceId":"claude-sonnet-5:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:28:30.636Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":38.3577,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Claude Sonnet 5 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"23c86e4a-c769-43c0-a056-79e3cd15834f","evaluatorModelSlug":"claude-sonnet-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"claude-sonnet-5:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":91.1111,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Sonnet 5 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"23c86e4a-c769-43c0-a056-79e3cd15834f","evaluatorModelSlug":"claude-sonnet-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"claude-sonnet-5:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":54.2824,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Sonnet 5 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"23c86e4a-c769-43c0-a056-79e3cd15834f","evaluatorModelSlug":"claude-sonnet-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"claude-sonnet-5:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":80.5243,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Sonnet 5 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"23c86e4a-c769-43c0-a056-79e3cd15834f","evaluatorModelSlug":"claude-sonnet-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"claude-sonnet-5:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":53},{"_id":"6a65fcb3fa5dacfbab007c83","name":"Leanstral 1.5","slug":"leanstral-15","apiIdentifier":"labs-leanstral-1-5","family":"Leanstral","version":"1.5","description":"Leanstral 1.5 is Mistral AI free Lean 4 formal proof-engineering model for automated theorem proving, autoformalization, and real-world code verification. Official Mistral docs list it as a Labs model with 119B total parameters, 6.5B active parameters, a 256k context window, 128k max output, $0 pricing, and the labs-leanstral-1-5 API identifier.","organizationSlug":"mistral","organizationName":"Mistral AI","contextWindow":256000,"maxOutputTokens":128000,"inputPricePerMillion":0,"outputPricePerMillion":0,"capabilities":["text","reasoning","coding","formal-verification","tool-use","structured-output"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-06-30T00:00:00.000Z","tags":["formal-verification","lean-4","theorem-proving","open-weights","coding"],"benchmarks":[{"source":"mistral","name":"miniF2F","score":100,"rawName":"miniF2F validation and test sets","sourceUrl":"https://mistral.ai/fr/news/leanstral-1-5/","sourceType":"official-release-benchmark","unit":"percent","scale":"0-100","modelVariant":"Leanstral 1.5","evidenceId":"mistral-leanstral-15-launch","confidence":"high","date":"2026-07-02T00:00:00.000Z","observedDate":"2026-07-26T00:00:00.000Z"},{"source":"mistral","name":"PutnamBench","score":587,"rawName":"PutnamBench problems solved","sourceUrl":"https://mistral.ai/fr/news/leanstral-1-5/","sourceType":"official-release-benchmark","unit":"problems solved","scale":"0-672","modelVariant":"Leanstral 1.5","evidenceId":"mistral-leanstral-15-launch","confidence":"high","date":"2026-07-02T00:00:00.000Z","observedDate":"2026-07-26T00:00:00.000Z"},{"source":"mistral","name":"FLTEval pass@8","score":43.2,"rawName":"FLTEval pass@8","sourceUrl":"https://mistral.ai/fr/news/leanstral-1-5/","sourceType":"official-release-benchmark","unit":"percent","scale":"0-100","modelVariant":"Leanstral 1.5","evidenceId":"mistral-leanstral-15-launch","confidence":"high","date":"2026-07-02T00:00:00.000Z","observedDate":"2026-07-26T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a36185efa46c955502c9c18","name":"LTX-2","slug":"ltx-2","apiIdentifier":"Lightricks/LTX-2","family":"LTX","version":"2","description":"LTX-2 is Lightricks’ audio-video generative model family with an official Python inference and LoRA trainer repository for builders experimenting with video generation workflows.","organizationSlug":"lightricks","organizationName":"Lightricks","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["video-generation","audio-generation","image-to-video","lora-training"],"category":"Video Model","current":true,"deprecated":false,"releaseDate":"2026-06-20T04:34:37.599Z","tags":["video-generation","audio-video","lora","diffusion","open-source"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a2b8d28e64a1d71fab92f93","name":"DiffusionGemma","slug":"diffusiongemma","apiIdentifier":"google/diffusiongemma-26b-a4b-it","family":"Gemma","version":"26B A4B","description":"DiffusionGemma is Google DeepMind’s experimental open-weights text-diffusion model based on Gemma 4 26B A4B. It uses discrete diffusion and parallel canvas denoising to trade some benchmark quality for much faster local generation on dedicated GPUs.","organizationSlug":"google","organizationName":"Google","contextWindow":256000,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","vision","code","reasoning","local-inference"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-06-10T00:00:00.000Z","tags":["diffusion","gemma","open-weights","local-inference","multimodal","mixture-of-experts"],"benchmarks":[{"source":"official-google-model-card","name":"MMLU Pro","score":77.6,"date":"2026-06-10T00:00:00.000Z"},{"source":"official-google-model-card","name":"GPQA Diamond","score":73.2,"date":"2026-06-10T00:00:00.000Z"},{"source":"official-google-model-card","name":"LiveCodeBench v6","score":69.1,"date":"2026-06-10T00:00:00.000Z"},{"source":"official-google-model-card","name":"MMMLU","score":81.5,"date":"2026-06-10T00:00:00.000Z"},{"source":"official-google-model-card","name":"HLE no tools","score":11,"date":"2026-06-10T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a689fd9eb09e1180791b5ec","name":"Claude Fable 5","slug":"claude-fable-5","apiIdentifier":"claude-fable-5","family":"Claude Fable","version":"5","description":"Claude Fable 5 is Anthropic's most capable widely released model for ambitious, long-horizon agentic work. Official Claude docs list the claude-fable-5 API ID, 1M token context window, 128k max output, always-on adaptive thinking, and $10 input / $50 output pricing per million tokens.","organizationSlug":"anthropic","organizationName":"Anthropic","contextWindow":1000000,"maxOutputTokens":128000,"inputPricePerMillion":10,"outputPricePerMillion":50,"capabilities":["text","vision","reasoning","coding","tool-use","agents","long-context"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-06-09T00:00:00.000Z","tags":["reasoning","coding","agentic","multimodal","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index","score":60,"rawName":"Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)","sourceUrl":"https://artificialanalysis.ai/leaderboards/models","sourceType":"benchmark-leaderboard","unit":"index","scale":"Artificial Analysis Intelligence Index","exactModelVariant":"Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)","evidenceId":"2026-07-28-artificial-analysis-leaderboard","confidence":"high","date":"2026-07-28T00:00:00.000Z","observedDate":"2026-07-28T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1.1","score":62,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1.1","sourceUrl":"https://artificialanalysis.ai/models/claude-fable-5","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)","exactModelVariant":"Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)","evidenceId":"sb-aa-claude-fable-5","confidence":"high","date":"2026-08-15T00:00:00.000Z","observedDate":"2026-08-15T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/claude-fable-5","name":"Artificial Analysis Intelligence Index v4.3","score":50,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-fable-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)","configuration":{"protocol":"Artificial Analysis Intelligence Index v4.3; exact effort/fallback in model variant"},"benchmarkVersion":"4.3","artifactId":"ceeaabf7e5ba792f","evidenceId":"claude-fable-5:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:28:32.627Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":49.6994,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd55210d-358e-4df1-ba9c-9acb5f186cc9","evaluatorModelSlug":"claude-fable-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"claude-fable-5:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":92.6263,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd55210d-358e-4df1-ba9c-9acb5f186cc9","evaluatorModelSlug":"claude-fable-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"claude-fable-5:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":60.9954,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd55210d-358e-4df1-ba9c-9acb5f186cc9","evaluatorModelSlug":"claude-fable-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"claude-fable-5:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":84.6442,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd55210d-358e-4df1-ba9c-9acb5f186cc9","evaluatorModelSlug":"claude-fable-5","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"claude-fable-5:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":62},{"_id":"6a747daf28c0324e4bd1a650","name":"North Mini Code","slug":"north-mini-code","apiIdentifier":"north-mini-code-1-0","family":"North","version":"1.0","description":"North Mini Code is Cohere's open-weight Mixture-of-Experts model for agentic coding, repository-level software engineering, terminal workflows, tool use, and local deployment.","organizationSlug":"cohere","organizationName":"Cohere","contextWindow":262144,"maxOutputTokens":65536,"inputPricePerMillion":0,"outputPricePerMillion":0,"capabilities":["reasoning","agentic-coding","tool-use","structured-output","multilingual","local-deployment"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-06-09T00:00:00.000Z","tags":["coding","reasoning","open-weights","tool-use","mixture-of-experts"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1","score":20,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1","sourceUrl":"https://artificialanalysis.ai/models/north-mini-code","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"North Mini Code (Reasoning)","evidenceId":"sb-aa-north-model","confidence":"high","date":"2026-08-06T00:00:00.000Z","observedDate":"2026-08-06T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":75.6566,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"North Mini Code","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"1e9907e0-ffac-4595-b006-962e4f1da7cf","evaluatorModelSlug":"north-mini-code","reasoning":true,"effort":null,"selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"north-mini-code:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":38.7731,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"North Mini Code","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"1e9907e0-ffac-4595-b006-962e4f1da7cf","evaluatorModelSlug":"north-mini-code","reasoning":true,"effort":null,"selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"north-mini-code:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":35.5805,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"North Mini Code","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"1e9907e0-ffac-4595-b006-962e4f1da7cf","evaluatorModelSlug":"north-mini-code","reasoning":true,"effort":null,"selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"north-mini-code:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":20},{"_id":"6a152270b06fb826885be13f","name":"DINOv3","slug":"dinov3","apiIdentifier":"facebookresearch/dinov3","family":"DINO","version":"v3","description":"DINOv3 is a Meta AI self-supervised vision foundation model family with public PyTorch implementation, pretrained backbones, model checkpoints, ImageNet classifier-head support, and dense visual features for computer-vision research.","organizationSlug":"meta","organizationName":"Meta","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["vision","embeddings","image-understanding","self-supervised-learning"],"category":"Vision Model","current":true,"deprecated":false,"releaseDate":"2026-05-26T04:32:44.708Z","tags":["vision-model","self-supervised-learning","computer-vision","meta-ai","pytorch","open-weights"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a0552a8fb24974d0528bd41","name":"Needle","slug":"needle","apiIdentifier":"Cactus-Compute/needle","family":"Needle","version":"26M","description":"Needle is an experimental 26 million parameter function-calling model from Cactus Compute. It is optimized for single-shot tool calls on small consumer devices and open weights are available on Hugging Face.","organizationSlug":"cactus-compute","organizationName":"Cactus Compute","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","tool-use","function-calling","on-device"],"category":"Small Language Model","current":true,"deprecated":false,"releaseDate":"2026-05-14T04:42:15.812Z","tags":["function-calling","tool-use","on-device-ai","open-weights","edge-ai"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69f02d92e623bffb718123a0","name":"GPT-5.5","slug":"gpt-55","apiIdentifier":"openai/gpt-5.5","family":"GPT","version":"5.5","description":"GPT-5.5 is OpenAI's smartest and most intuitive model, built for agentic work like coding, research, and data analysis. It matches GPT-5.4 per-token latency while delivering higher intelligence with significantly fewer tokens. Supports a 1,050,000 token context window and five reasoning effort levels (none through xhigh).","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":5,"outputPricePerMillion":30,"capabilities":["text","vision","code","tool-use","extended-thinking","computer-use","web-search"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-04-23T00:00:00.000Z","tags":["reasoning","coding","agentic","long-context","multimodal","tool-use"],"benchmarks":[{"source":"openai","name":"MMLU","score":92.4,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"GPQA Diamond","score":93.6,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"ARC-AGI-2","score":85,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"Terminal-Bench 2.0","score":82.7,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"SWE-bench Pro","score":58.6,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"OSWorld-Verified","score":78.7,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"BrowseComp","score":84.4,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"MMMU-Pro","score":81.2,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"FrontierMath Tier 4","score":35.4,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"HLE (with tools)","score":52.2,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"GDPval","score":84.9,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"Toolathlon","score":55.6,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"CyberGym","score":81.8,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"MRCR v2 512K-1M","score":74,"date":"2026-04-23T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":38.6292,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GPT-5.5 (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"1f054429-397e-4fdb-9e71-67bc92c1735e","evaluatorModelSlug":"gpt-5-5","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gpt-55:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":93.5354,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.5 (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"1f054429-397e-4fdb-9e71-67bc92c1735e","evaluatorModelSlug":"gpt-5-5","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-55:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":55.787,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.5 (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"1f054429-397e-4fdb-9e71-67bc92c1735e","evaluatorModelSlug":"gpt-5-5","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-55:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":84.2697,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.5 (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"1f054429-397e-4fdb-9e71-67bc92c1735e","evaluatorModelSlug":"gpt-5-5","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gpt-55:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1504,"arenaVotes":0,"aaIntelligenceIndex":60},{"_id":"69f02d92e623bffb718123a1","name":"GPT-5.5 Pro","slug":"gpt-55-pro","apiIdentifier":"openai/gpt-5.5-pro","family":"GPT","version":"5.5 Pro","description":"GPT-5.5 Pro uses more compute to think harder and provide consistently better answers than GPT-5.5. Designed for the toughest problems, some requests may take several minutes. It excels at FrontierMath Tier 4 (39.6%) and BrowseComp (90.1%), offering the highest intelligence available from OpenAI at a premium price point.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":30,"outputPricePerMillion":180,"capabilities":["text","vision","code","tool-use","extended-thinking","web-search"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-04-23T00:00:00.000Z","tags":["reasoning","coding","agentic","long-context","multimodal","pro","tool-use"],"benchmarks":[{"source":"openai","name":"FrontierMath Tier 4","score":39.6,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"FrontierMath Tier 1-3","score":52.4,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"HLE (with tools)","score":57.2,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"HLE (no tools)","score":43.1,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"BrowseComp","score":90.1,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"GeneBench","score":33.2,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"GDPval","score":82.3,"date":"2026-04-23T00:00:00.000Z"},{"source":"openai","name":"Investment Banking Modeling","score":88.6,"date":"2026-04-23T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6b3","name":"Claude Opus 4.7","slug":"claude-opus-47","apiIdentifier":"anthropic/claude-opus-4.7","family":"Claude","version":"4.7","description":"Claude Opus 4.7 is Anthropic's most capable generally available model, with significant improvements in advanced software engineering, agentic tool use, and vision resolution. Achieves 87.6% on SWE-bench Verified and 94.2% on GPQA Diamond. Supports up to 1,000,000 token context window with 3.3x higher-resolution vision than Opus 4.6.","organizationSlug":"anthropic","organizationName":"Anthropic","contextWindow":1000000,"maxOutputTokens":128000,"inputPricePerMillion":5,"outputPricePerMillion":25,"capabilities":["text","vision","code","tool-use"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-04-16T00:00:00.000Z","tags":["multimodal","long-context"],"benchmarks":[{"source":"anthropic","name":"MMLU","score":84.7,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"MMLU-Pro","score":78.1,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"MMMLU","score":92,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"GPQA Diamond","score":94.2,"date":"2026-04-16T00:00:00.000Z"},{"source":"artificial-analysis","name":"HLE","score":54.7,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"SWE-bench Verified","score":87.6,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"SWE-bench Pro","score":64.3,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"SWE-bench Multilingual+Multimodal","score":80.5,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"Terminal-Bench","score":69.4,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"MCP-Atlas","score":77.3,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"Berkeley Function Calling","score":77.3,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"OSWorld-Verified","score":78,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"BrowseComp","score":79.3,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"CharXiv-R","score":91,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"DocVQA","score":93.1,"date":"2026-04-16T00:00:00.000Z"},{"source":"anthropic","name":"CyberGym","score":73.1,"date":"2026-04-16T00:00:00.000Z"},{"source":"artificial-analysis","name":"GDPVal-AA Elo","score":1753,"date":"2026-04-16T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":91.4141,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Opus 4.7 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"e9a09db3-8fd6-41dd-ba2f-20e0a2bff7f2","evaluatorModelSlug":"claude-opus-4-7","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"claude-opus-47:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":83.1461,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Opus 4.7 (Adaptive Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"e9a09db3-8fd6-41dd-ba2f-20e0a2bff7f2","evaluatorModelSlug":"claude-opus-4-7","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"claude-opus-47:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1504,"arenaVotes":3485,"aaIntelligenceIndex":57},{"_id":"69e45e9faccbb529d120a6b5","name":"GLM 5.1","slug":"glm-51","apiIdentifier":"z-ai/glm-5.1","family":"GLM","version":"5.1","description":"GLM 5.1 is a large language model from Z.ai. Supports up to 202,752 token context window. Available from $0.95/M input tokens.","organizationSlug":"z-ai","organizationName":"Z.ai","contextWindow":202752,"maxOutputTokens":65535,"inputPricePerMillion":0.95,"outputPricePerMillion":3.15,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-04-07T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"https://huggingface.co/zai-org/GLM-5.1","name":"GPQA Diamond","score":86.2,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/zai-org/GLM-5.1","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"GLM-5.1","configuration":{"protocol":"GLM-5.1 author-reported evaluation"},"artifactId":"b6bc18a847465675","evidenceId":"glm-51:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:03.981Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":26.4455,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GLM-5.1 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"5aa1c578-af76-4b91-8699-cdd43582b3af","evaluatorModelSlug":"glm-5-1","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"glm-51:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":86.7677,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5.1 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"5aa1c578-af76-4b91-8699-cdd43582b3af","evaluatorModelSlug":"glm-5-1","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"glm-51:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":44.7917,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5.1 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"5aa1c578-af76-4b91-8699-cdd43582b3af","evaluatorModelSlug":"glm-5-1","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"glm-51:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":61.7978,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5.1 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"5aa1c578-af76-4b91-8699-cdd43582b3af","evaluatorModelSlug":"glm-5-1","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"glm-51:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1472,"arenaVotes":7179,"aaIntelligenceIndex":51},{"_id":"69e45e9faccbb529d120a6d1","name":"Gemma 4 26B A4B","slug":"gemma-4-26b-a4b-google","apiIdentifier":"google/gemma-4-26b-a4b-it","family":"Gemma","version":"26B A4B","description":"Gemma 4 26B A4B is a multimodal llm from Google. Supports up to 262,144 token context window. Available from $0.08/M input tokens.","organizationSlug":"google","organizationName":"Google","contextWindow":262144,"maxOutputTokens":null,"inputPricePerMillion":0.08,"outputPricePerMillion":0.35,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-04-03T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":79.1919,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemma 4 26B A4B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"6e1b44ff-c227-496b-aef4-19b70cd18c76","evaluatorModelSlug":"gemma-4-26b-a4b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemma-4-26b-a4b-google:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":38.9513,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemma 4 26B A4B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"6e1b44ff-c227-496b-aef4-19b70cd18c76","evaluatorModelSlug":"gemma-4-26b-a4b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gemma-4-26b-a4b-google:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1439,"arenaVotes":5782,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6b1","name":"Gemma 4 26B A4B","slug":"gemma-4-26b-a4b","apiIdentifier":"google/gemma-4-26b-a4b-it:free","family":"Gemma","version":"26B A4B","description":"Gemma 4 26B A4B is a multimodal llm from Google. Supports up to 262,144 token context window. Available free with rate limits on OpenRouter.","organizationSlug":"google","organizationName":"Google","contextWindow":262144,"maxOutputTokens":32768,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-04-03T00:00:00.000Z","tags":["multimodal","video","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":79.1919,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemma 4 26B A4B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"6e1b44ff-c227-496b-aef4-19b70cd18c76","evaluatorModelSlug":"gemma-4-26b-a4b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemma-4-26b-a4b:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":38.9513,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemma 4 26B A4B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"6e1b44ff-c227-496b-aef4-19b70cd18c76","evaluatorModelSlug":"gemma-4-26b-a4b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gemma-4-26b-a4b:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1439,"arenaVotes":5782,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6eb","name":"Gemma 4 31B","slug":"gemma-4-31b","apiIdentifier":"google/gemma-4-31b-it:free","family":"Gemma","version":"31B","description":"Gemma 4 31B is a multimodal llm from Google. Supports up to 262,144 token context window. Available free with rate limits on OpenRouter.","organizationSlug":"google","organizationName":"Google","contextWindow":262144,"maxOutputTokens":32768,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-04-02T00:00:00.000Z","tags":["multimodal","video","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":15.4157,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Gemma 4 31B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd26a386-4873-46ff-b853-d239050025a2","evaluatorModelSlug":"gemma-4-31b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gemma-4-31b:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":85.6566,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemma 4 31B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd26a386-4873-46ff-b853-d239050025a2","evaluatorModelSlug":"gemma-4-31b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemma-4-31b:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":45.4861,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemma 4 31B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd26a386-4873-46ff-b853-d239050025a2","evaluatorModelSlug":"gemma-4-31b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemma-4-31b:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":43.4457,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemma 4 31B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd26a386-4873-46ff-b853-d239050025a2","evaluatorModelSlug":"gemma-4-31b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gemma-4-31b:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1451,"arenaVotes":5827,"aaIntelligenceIndex":39},{"_id":"69e45e9faccbb529d120a701","name":"Gemma 4 31B","slug":"gemma-4-31b-google","apiIdentifier":"google/gemma-4-31b-it","family":"Gemma","version":"31B","description":"Gemma 4 31B is a multimodal llm from Google. Supports up to 262,144 token context window. Available from $0.13/M input tokens.","organizationSlug":"google","organizationName":"Google","contextWindow":262144,"maxOutputTokens":null,"inputPricePerMillion":0.13,"outputPricePerMillion":0.38,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-04-02T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":15.4157,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Gemma 4 31B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd26a386-4873-46ff-b853-d239050025a2","evaluatorModelSlug":"gemma-4-31b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gemma-4-31b-google:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":85.6566,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemma 4 31B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd26a386-4873-46ff-b853-d239050025a2","evaluatorModelSlug":"gemma-4-31b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemma-4-31b-google:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":45.4861,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemma 4 31B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd26a386-4873-46ff-b853-d239050025a2","evaluatorModelSlug":"gemma-4-31b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gemma-4-31b-google:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":43.4457,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Gemma 4 31B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"cd26a386-4873-46ff-b853-d239050025a2","evaluatorModelSlug":"gemma-4-31b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gemma-4-31b-google:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1451,"arenaVotes":5827,"aaIntelligenceIndex":39},{"_id":"69e45e9faccbb529d120a6b0","name":"Qwen3.6 Plus","slug":"qwen36-plus","apiIdentifier":"qwen/qwen3.6-plus","family":"Qwen","version":null,"description":"Qwen3.6 Plus is a multimodal llm from Alibaba. Supports up to 1,000,000 token context window. Achieves 87.5% on MMLU. Available from $0.33/M input tokens.","organizationSlug":"alibaba","organizationName":"Alibaba","contextWindow":1000000,"maxOutputTokens":65536,"inputPricePerMillion":0.325,"outputPricePerMillion":1.95,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-04-02T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[{"source":"alibaba","name":"GPQA","score":90.4,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"LiveCodeBench v6","score":87.1,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"SWE-bench Verified","score":78.8,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"Terminal-Bench 2.0","score":61.6,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"TAU3-Bench","score":70.7,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"VideoMME","score":87.8,"date":"2026-04-28T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":88.1818,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Qwen3.6 Plus","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"f371ad68-6947-4767-a78f-1f6c81f96b93","evaluatorModelSlug":"qwen3-6-plus","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"qwen36-plus:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":61.4232,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Qwen3.6 Plus","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"f371ad68-6947-4767-a78f-1f6c81f96b93","evaluatorModelSlug":"qwen3-6-plus","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"qwen36-plus:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1467,"arenaVotes":9591,"aaIntelligenceIndex":50},{"_id":"69e45e9faccbb529d120a6ba","name":"Trinity Large Thinking","slug":"trinity-large-thinking","apiIdentifier":"arcee-ai/trinity-large-thinking","family":"Trinity","version":null,"description":"Trinity Large Thinking is a large language model from Arcee AI. Supports up to 262,144 token context window. Available from $0.22/M input tokens.","organizationSlug":"arcee-ai","organizationName":"Arcee AI","contextWindow":262144,"maxOutputTokens":262144,"inputPricePerMillion":0.22,"outputPricePerMillion":0.85,"capabilities":["text","code","extended-thinking"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-04-01T00:00:00.000Z","tags":["reasoning","budget-friendly","long-context","chain-of-thought"],"benchmarks":[{"source":"https://huggingface.co/arcee-ai/Trinity-Large-Thinking","name":"GPQA Diamond","score":76.3,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/arcee-ai/Trinity-Large-Thinking","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Trinity-Large-Thinking","configuration":{"protocol":"Trinity-Large-Thinking author-reported evaluation"},"artifactId":"3400e47fee8b0564","evidenceId":"trinity-large-thinking:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:25:26.958Z"},{"source":"https://huggingface.co/arcee-ai/Trinity-Large-Thinking","name":"MMLU-Pro","score":83.4,"canonicalName":"MMLU-Pro","rawName":"MMLU-Pro","sourceUrl":"https://huggingface.co/arcee-ai/Trinity-Large-Thinking","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Trinity-Large-Thinking","configuration":{"protocol":"Trinity-Large-Thinking author-reported evaluation"},"artifactId":"3400e47fee8b0564","evidenceId":"trinity-large-thinking:new-benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:25:26.958Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":10.8821,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Trinity Large Thinking","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b6d2e43d-3082-43f5-9318-0f4dbcb54163","evaluatorModelSlug":"trinity-large-thinking","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"trinity-large-thinking:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":75.1515,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Trinity Large Thinking","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b6d2e43d-3082-43f5-9318-0f4dbcb54163","evaluatorModelSlug":"trinity-large-thinking","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"trinity-large-thinking:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":40.625,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Trinity Large Thinking","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b6d2e43d-3082-43f5-9318-0f4dbcb54163","evaluatorModelSlug":"trinity-large-thinking","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"trinity-large-thinking:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":20.5993,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Trinity Large Thinking","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b6d2e43d-3082-43f5-9318-0f4dbcb54163","evaluatorModelSlug":"trinity-large-thinking","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"trinity-large-thinking:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1376,"arenaVotes":15015,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6d5","name":"GLM 5V Turbo","slug":"glm-5v-turbo","apiIdentifier":"z-ai/glm-5v-turbo","family":"GLM","version":"5V Turbo","description":"GLM 5V Turbo is a multimodal llm from Z.ai. Supports up to 202,752 token context window. Available from $1.20/M input tokens.","organizationSlug":"z-ai","organizationName":"Z.ai","contextWindow":202752,"maxOutputTokens":131072,"inputPricePerMillion":1.2,"outputPricePerMillion":4,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-04-01T00:00:00.000Z","tags":["multimodal","video","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":80.9091,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM 5V Turbo (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"248deb0d-426c-4fa9-86fa-bc60aa9c3719","evaluatorModelSlug":"glm-5v-turbo","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"glm-5v-turbo:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1378,"arenaVotes":2807,"aaIntelligenceIndex":null},{"_id":"6a1e5e6e2ae589a72aa3c737","name":"VoxCPM","slug":"voxcpm","apiIdentifier":"openbmb/VoxCPM2","family":"VoxCPM","version":"2","description":"VoxCPM is OpenBMB’s tokenizer-free text-to-speech model family. VoxCPM2 is a 2B-parameter release for multilingual speech generation, voice design, and controllable voice cloning.","organizationSlug":"openbmb","organizationName":"OpenBMB","contextWindow":4096,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text-to-speech","multilingual","voice-cloning","voice-design","streaming-inference"],"category":"Audio Model","current":true,"deprecated":false,"releaseDate":"2026-04-01T00:00:00.000Z","tags":["text-to-speech","voice-cloning","multilingual","speech-generation","open-source"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6b7","name":"Grok 4.20 Multi-Agent","slug":"grok-420-multi-agent","apiIdentifier":"x-ai/grok-4.20-multi-agent","family":"Grok","version":"Multi-Agent","description":"Grok 4.20 Multi-Agent is a multimodal llm from xAI. Supports up to 2,000,000 token context window. Available from $2.00/M input tokens.","organizationSlug":"xai","organizationName":"xAI","contextWindow":2000000,"maxOutputTokens":null,"inputPricePerMillion":2,"outputPricePerMillion":6,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-03-31T00:00:00.000Z","tags":["multimodal","long-context"],"benchmarks":[],"arenaScore":1476,"arenaVotes":12049,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6d7","name":"Grok 4.20","slug":"grok-420","apiIdentifier":"x-ai/grok-4.20","family":"Grok","version":null,"description":"Grok 4.20 is a multimodal llm from xAI. Supports up to 2,000,000 token context window. Available from $2.00/M input tokens.","organizationSlug":"xai","organizationName":"xAI","contextWindow":2000000,"maxOutputTokens":null,"inputPricePerMillion":2,"outputPricePerMillion":6,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-03-31T00:00:00.000Z","tags":["multimodal","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":91.1111,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Grok 4.20 0309 v2 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"c72cb85a-18a4-4235-b455-77dff2f16c50","evaluatorModelSlug":"grok-4-20","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"grok-420:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1485,"arenaVotes":12069,"aaIntelligenceIndex":49},{"_id":"6a9f9023c2731fac11c1dced","name":"OmniVoice","slug":"omnivoice","apiIdentifier":"k2-fsa/OmniVoice","family":"OmniVoice","version":"2026","description":"OmniVoice is an open-source massively multilingual zero-shot text-to-speech and voice-cloning model from K2-FSA. The project README says it supports more than 600 languages, voice design controls, and fast inference with real-time factor as low as 0.025.","organizationSlug":"k2-fsa","organizationName":"K2-FSA","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text-to-speech","voice-cloning","multilingual-audio","voice-design"],"category":"Audio Model","current":true,"deprecated":false,"releaseDate":"2026-03-31T00:00:00.000Z","tags":["text-to-speech","voice-cloning","multilingual","speech-ai","open-source","diffusion-model"],"benchmarks":[{"source":"official-github-readme","name":"Real-time factor (best-case author claim)","score":0.025,"canonicalName":"Real-time factor","rawName":"RTF as low as 0.025","sourceUrl":"https://github.com/k2-fsa/OmniVoice","sourceType":"first-party-model-card","unit":"ratio","scale":"nonnegative","modelVariant":"k2-fsa/OmniVoice","configuration":{"claim":"Best-case advertised RTF; hardware and batch unspecified in this headline"},"artifactId":"6242800a9b7187f9","evidenceId":"omnivoice:benchmark:0","confidence":"high","higherIsBetter":false,"comparisonEligible":false,"exclusionReason":"Headline best-case claim lacks matching hardware and batch conditions","date":"2026-09-08T04:36:28.449Z","observedAt":"2026-09-12T16:18:03.980Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6c5","name":"Reka Edge","slug":"reka-edge","apiIdentifier":"rekaai/reka-edge","family":"Reka","version":null,"description":"Reka Edge is a multimodal llm from Reka. Supports up to 16,384 token context window. Available from $0.10/M input tokens.","organizationSlug":"reka","organizationName":"Reka","contextWindow":16384,"maxOutputTokens":16384,"inputPricePerMillion":0.1,"outputPricePerMillion":0.1,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-03-20T00:00:00.000Z","tags":["multimodal","video","budget-friendly"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6b9","name":"MiniMax M2.7","slug":"minimax-m27","apiIdentifier":"minimax/minimax-m2.7","family":"MiniMax","version":"M2.7","description":"MiniMax M2.7 is a large language model from MiniMax. Supports up to 196,608 token context window. Available from $0.30/M input tokens.","organizationSlug":"minimax","organizationName":"MiniMax","contextWindow":196608,"maxOutputTokens":null,"inputPricePerMillion":0.3,"outputPricePerMillion":1.2,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-03-18T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"https://huggingface.co/MiniMaxAI/MiniMax-M2.7","name":"SWE-bench Pro","score":56.22,"canonicalName":"SWE-bench Pro","rawName":"SWE-bench Pro","sourceUrl":"https://huggingface.co/MiniMaxAI/MiniMax-M2.7","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"MiniMax M2.7","configuration":{"protocol":"MiniMax release evaluation; harness and task subset must match the card"},"artifactId":"6464c0d9e9781bc5","evidenceId":"minimax-m27:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:09.750Z"},{"source":"https://huggingface.co/MiniMaxAI/MiniMax-M2.7","name":"Terminal-Bench 2","score":57,"canonicalName":"Terminal-Bench 2","rawName":"Terminal-Bench 2","sourceUrl":"https://huggingface.co/MiniMaxAI/MiniMax-M2.7","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"MiniMax M2.7","configuration":{"protocol":"MiniMax release evaluation; harness must match the card"},"benchmarkVersion":"2","artifactId":"6464c0d9e9781bc5","evidenceId":"minimax-m27:new-benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:09.750Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":23.2171,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"MiniMax-M2.7","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"4bbceacb-cf47-464b-b60f-e1d1fe016d67","evaluatorModelSlug":"minimax-m2-7","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"minimax-m27:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":87.3737,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"MiniMax-M2.7","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"4bbceacb-cf47-464b-b60f-e1d1fe016d67","evaluatorModelSlug":"minimax-m2-7","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"minimax-m27:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":50.1157,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"MiniMax-M2.7","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"4bbceacb-cf47-464b-b60f-e1d1fe016d67","evaluatorModelSlug":"minimax-m2-7","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"minimax-m27:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":55.4307,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"MiniMax-M2.7","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"4bbceacb-cf47-464b-b60f-e1d1fe016d67","evaluatorModelSlug":"minimax-m2-7","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"minimax-m27:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1403,"arenaVotes":8484,"aaIntelligenceIndex":50},{"_id":"69e45e9faccbb529d120a6cf","name":"GPT-5.4 Mini","slug":"gpt-54-mini","apiIdentifier":"openai/gpt-5.4-mini","family":"GPT","version":"Mini","description":"GPT-5.4 Mini is a multimodal llm from OpenAI. Supports up to 400,000 token context window. Available from $0.75/M input tokens.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":400000,"maxOutputTokens":128000,"inputPricePerMillion":0.75,"outputPricePerMillion":4.5,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-03-17T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[{"source":"openai","name":"GPQA Diamond","score":88,"date":"2026-03-16T00:00:00.000Z"},{"source":"openai","name":"SWE-bench Pro","score":54.4,"date":"2026-03-16T00:00:00.000Z"},{"source":"openai","name":"OSWorld-Verified","score":72.1,"date":"2026-03-16T00:00:00.000Z"},{"source":"openai","name":"Terminal-Bench 2.0","score":60,"date":"2026-03-16T00:00:00.000Z"},{"source":"openai","name":"MMMU-Pro","score":76.6,"date":"2026-03-16T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":24.5789,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GPT-5.4 mini (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"ba242e40-83b7-4cd3-a0e0-b56237984914","evaluatorModelSlug":"gpt-5-4-mini","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gpt-54-mini:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":87.4747,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.4 mini (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"ba242e40-83b7-4cd3-a0e0-b56237984914","evaluatorModelSlug":"gpt-5-4-mini","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-54-mini:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":52.0833,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.4 mini (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"ba242e40-83b7-4cd3-a0e0-b56237984914","evaluatorModelSlug":"gpt-5-4-mini","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-54-mini:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":59.176,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.4 mini (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"ba242e40-83b7-4cd3-a0e0-b56237984914","evaluatorModelSlug":"gpt-5-4-mini","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gpt-54-mini:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1458,"arenaVotes":9135,"aaIntelligenceIndex":48},{"_id":"69e45e9faccbb529d120a6af","name":"GPT-5.4 Nano","slug":"gpt-54-nano","apiIdentifier":"openai/gpt-5.4-nano","family":"GPT","version":"Nano","description":"GPT-5.4 Nano is a multimodal llm from OpenAI. Supports up to 400,000 token context window. Available from $0.20/M input tokens.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":400000,"maxOutputTokens":128000,"inputPricePerMillion":0.2,"outputPricePerMillion":1.25,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-03-17T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":21.2442,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GPT-5.4 nano (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d4fc3f33-f2b0-4da1-88ee-f1f82bd4de31","evaluatorModelSlug":"gpt-5-4-nano","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gpt-54-nano:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":81.7172,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.4 nano (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d4fc3f33-f2b0-4da1-88ee-f1f82bd4de31","evaluatorModelSlug":"gpt-5-4-nano","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-54-nano:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":47.2222,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.4 nano (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d4fc3f33-f2b0-4da1-88ee-f1f82bd4de31","evaluatorModelSlug":"gpt-5-4-nano","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-54-nano:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":60.6742,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.4 nano (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"d4fc3f33-f2b0-4da1-88ee-f1f82bd4de31","evaluatorModelSlug":"gpt-5-4-nano","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gpt-54-nano:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1403,"arenaVotes":8430,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6b2","name":"Mistral Small 4","slug":"mistral-small-4","apiIdentifier":"mistralai/mistral-small-2603","family":"Mistral","version":"4","description":"Mistral Small 4 is Mistral's Apache-2.0 open-weight hybrid model for instruction following, reasoning, and coding. It has 119B total parameters, 6.5B active parameters, and a 256K-token context window.","organizationSlug":"mistral","organizationName":"Mistral AI","contextWindow":262144,"maxOutputTokens":null,"inputPricePerMillion":0.15,"outputPricePerMillion":0.6,"capabilities":["text","vision","code","tool-use"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-03-16T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[{"source":"mistral","name":"MMLU-Pro","score":68.5,"date":"2025-03-05T00:00:00.000Z"},{"source":"mistral","name":"HumanEval","score":78.5,"date":"2025-03-05T00:00:00.000Z"},{"source":"mistral","name":"MGSM","score":72.3,"date":"2025-03-05T00:00:00.000Z"},{"source":"mistral","name":"GPQA Diamond","score":58.8,"date":"2025-03-05T00:00:00.000Z"},{"source":"mistral","name":"LiveCodeBench","score":64,"date":"2025-03-05T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":11.4511,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Mistral Small 4 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"3fd96175-4ef1-434c-8795-f873aec2abc1","evaluatorModelSlug":"mistral-small-4","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"mistral-small-4:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":76.8687,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Mistral Small 4 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"3fd96175-4ef1-434c-8795-f873aec2abc1","evaluatorModelSlug":"mistral-small-4","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"mistral-small-4:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":38.7731,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Mistral Small 4 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"3fd96175-4ef1-434c-8795-f873aec2abc1","evaluatorModelSlug":"mistral-small-4","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"mistral-small-4:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":20.9738,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Mistral Small 4 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"3fd96175-4ef1-434c-8795-f873aec2abc1","evaluatorModelSlug":"mistral-small-4","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"mistral-small-4:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1357,"arenaVotes":17729,"aaIntelligenceIndex":28},{"_id":"69e45e9faccbb529d120a6ef","name":"GLM 5 Turbo","slug":"glm-5-turbo","apiIdentifier":"z-ai/glm-5-turbo","family":"GLM","version":"5 Turbo","description":"GLM 5 Turbo is a large language model from Z.ai. Supports up to 202,752 token context window. Available from $1.20/M input tokens.","organizationSlug":"z-ai","organizationName":"Z.ai","contextWindow":202752,"maxOutputTokens":131072,"inputPricePerMillion":1.2,"outputPricePerMillion":4,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-03-15T00:00:00.000Z","tags":["long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":84.7475,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5-Turbo","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"149096f3-57b7-4413-80c2-a2c010a2995a","evaluatorModelSlug":"glm-5-turbo","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"glm-5-turbo:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1456,"arenaVotes":15839,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6b8","name":"Nemotron 3 Super","slug":"nemotron-3-super","apiIdentifier":"nvidia/nemotron-3-super-120b-a12b:free","family":"Nemotron","version":null,"description":"Nemotron 3 Super is NVIDIA's 120B-parameter model with 12B active parameters. Its default Hugging Face configuration uses 256K context; NVIDIA documents up to 1M model context with additional VRAM. Hosted endpoint limits may differ.","organizationSlug":"nvidia","organizationName":"NVIDIA","contextWindow":262144,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-03-11T00:00:00.000Z","tags":["long-context"],"benchmarks":[],"arenaScore":1361,"arenaVotes":7413,"aaIntelligenceIndex":36},{"_id":"69e45e9faccbb529d120a6d8","name":"Nemotron 3 Super","slug":"nemotron-3-super-nvidia","apiIdentifier":"nvidia/nemotron-3-super-120b-a12b","family":"Nemotron","version":null,"description":"Nemotron 3 Super is a large language model from NVIDIA. Supports up to 262,144 token context window. Available from $0.09/M input tokens.","organizationSlug":"nvidia","organizationName":"NVIDIA","contextWindow":262144,"maxOutputTokens":null,"inputPricePerMillion":0.09,"outputPricePerMillion":0.45,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-03-11T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[],"arenaScore":1361,"arenaVotes":7413,"aaIntelligenceIndex":36},{"_id":"69e45e9faccbb529d120a6c0","name":"Seed-2.0-Lite","slug":"seed-20-lite","apiIdentifier":"bytedance-seed/seed-2.0-lite","family":"Seed-2.0-Lite","version":null,"description":"Seed-2.0-Lite is a multimodal llm from ByteDance. Supports up to 262,144 token context window. Available from $0.25/M input tokens.","organizationSlug":"bytedance","organizationName":"ByteDance","contextWindow":262144,"maxOutputTokens":131072,"inputPricePerMillion":0.25,"outputPricePerMillion":2,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-03-10T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6d0","name":"Qwen3.5-9B","slug":"qwen35-9b","apiIdentifier":"qwen/qwen3.5-9b","family":"Qwen","version":null,"description":"Qwen3.5-9B is a multimodal llm from Alibaba. Supports up to 262,144 token context window. Achieves 87.5% on MMLU. Available from $0.10/M input tokens.","organizationSlug":"alibaba","organizationName":"Alibaba","contextWindow":262144,"maxOutputTokens":null,"inputPricePerMillion":0.1,"outputPricePerMillion":0.15,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-03-10T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[{"source":"alibaba","name":"MMLU-Pro","score":82.5,"canonicalName":"MMLU-Pro","rawName":"MMLU-Pro","sourceUrl":"https://huggingface.co/Qwen/Qwen3.5-9B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5-9B","configuration":{"protocol":"Qwen3.5 author-reported model-card evaluation; consult linked evaluation settings"},"artifactId":"e0a64484787e9317","evidenceId":"qwen35-9b:benchmark:0","confidence":"high","higherIsBetter":true,"date":"2026-04-28T00:00:00.000Z","observedAt":"2026-09-12T16:25:35.960Z"},{"source":"alibaba","name":"GPQA Diamond","score":81.7,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/Qwen/Qwen3.5-9B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5-9B","configuration":{"protocol":"Qwen3.5 author-reported model-card evaluation; consult linked evaluation settings"},"artifactId":"e0a64484787e9317","evidenceId":"qwen35-9b:benchmark:1","confidence":"high","higherIsBetter":true,"date":"2026-04-28T00:00:00.000Z","observedAt":"2026-09-12T16:25:35.960Z"},{"source":"alibaba","name":"IFEval","score":91.5,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"HMMT 2025","score":83.2,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"TAU2-Bench","score":79.1,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"MMMU-Pro","score":70.1,"date":"2026-04-28T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":80.6061,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5 9B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"2d28a13a-096e-475a-beb8-26bbd1c7d51c","evaluatorModelSlug":"qwen3-5-9b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"qwen35-9b:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":29.2135,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5 9B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"2d28a13a-096e-475a-beb8-26bbd1c7d51c","evaluatorModelSlug":"qwen3-5-9b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"qwen35-9b:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1399,"arenaVotes":14459,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6ff","name":"GPT-5.4","slug":"gpt-54","apiIdentifier":"openai/gpt-5.4","family":"GPT","version":null,"description":"GPT-5.4 is a multimodal llm from OpenAI. Supports up to 1,050,000 token context window. Available from $2.50/M input tokens.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":2.5,"outputPricePerMillion":15,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-03-05T00:00:00.000Z","tags":["multimodal","long-context"],"benchmarks":[{"source":"openai","name":"GPQA Diamond","score":93,"date":"2026-03-05T00:00:00.000Z"},{"source":"openai","name":"SWE-bench Pro","score":57.7,"date":"2026-03-05T00:00:00.000Z"},{"source":"openai","name":"OSWorld-Verified","score":75,"date":"2026-03-05T00:00:00.000Z"},{"source":"openai","name":"Terminal-Bench 2.0","score":75.1,"date":"2026-03-05T00:00:00.000Z"},{"source":"openai","name":"ARC-AGI-2","score":73.3,"date":"2026-03-05T00:00:00.000Z"},{"source":"openai","name":"BrowseComp","score":82.7,"date":"2026-03-05T00:00:00.000Z"},{"source":"openai","name":"MMMU-Pro","score":81.2,"date":"2026-03-05T00:00:00.000Z"},{"source":"openai","name":"MATH-500","score":94.6,"date":"2026-03-05T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":92.0202,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.4 (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"a89c4b28-2d8c-456e-88ea-255fb51fd2b6","evaluatorModelSlug":"gpt-5-4","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-54:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":78.2772,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-5.4 (xhigh)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"a89c4b28-2d8c-456e-88ea-255fb51fd2b6","evaluatorModelSlug":"gpt-5-4","reasoning":true,"effort":"xhigh","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gpt-54:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1468,"arenaVotes":11988,"aaIntelligenceIndex":57},{"_id":"69e45e9faccbb529d120a6e9","name":"GPT-5.4 Pro","slug":"gpt-54-pro","apiIdentifier":"openai/gpt-5.4-pro","family":"GPT","version":"Pro","description":"GPT-5.4 Pro is a multimodal llm from OpenAI. Supports up to 1,050,000 token context window. Available from $30.00/M input tokens.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":30,"outputPricePerMillion":180,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-03-05T00:00:00.000Z","tags":["multimodal","long-context"],"benchmarks":[{"source":"openai","name":"GPQA Diamond","score":93,"date":"2026-03-05T00:00:00.000Z"},{"source":"openai","name":"SWE-bench Pro","score":57.7,"date":"2026-03-05T00:00:00.000Z"},{"source":"openai","name":"MATH-500","score":94.6,"date":"2026-03-05T00:00:00.000Z"}],"arenaScore":1482,"arenaVotes":11568,"aaIntelligenceIndex":57},{"_id":"69e45e9faccbb529d120a6e0","name":"Seed-2.0-Mini","slug":"seed-20-mini","apiIdentifier":"bytedance-seed/seed-2.0-mini","family":"Seed-2.0-Mini","version":null,"description":"Seed-2.0-Mini is a multimodal llm from ByteDance. Supports up to 262,144 token context window. Available from $0.10/M input tokens.","organizationSlug":"bytedance","organizationName":"ByteDance","contextWindow":262144,"maxOutputTokens":131072,"inputPricePerMillion":0.1,"outputPricePerMillion":0.4,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-02-26T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a700","name":"Qwen3.5-27B","slug":"qwen35-27b","apiIdentifier":"qwen/qwen3.5-27b","family":"Qwen","version":null,"description":"Qwen3.5-27B is a multimodal llm from Alibaba. Supports up to 262,144 token context window. Achieves 87.5% on MMLU. Available from $0.20/M input tokens.","organizationSlug":"alibaba","organizationName":"Alibaba","contextWindow":262144,"maxOutputTokens":65536,"inputPricePerMillion":0.195,"outputPricePerMillion":1.56,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-02-25T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[{"source":"alibaba","name":"MMLU-Pro","score":86.1,"canonicalName":"MMLU-Pro","rawName":"MMLU-Pro","sourceUrl":"https://huggingface.co/Qwen/Qwen3.5-27B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5-27B","configuration":{"protocol":"Qwen3.5 author-reported model-card evaluation; consult linked evaluation settings"},"artifactId":"76e97b66c81294a6","evidenceId":"qwen35-27b:benchmark:0","confidence":"high","higherIsBetter":true,"date":"2026-04-28T00:00:00.000Z","observedAt":"2026-09-12T16:17:39.480Z"},{"source":"alibaba","name":"GPQA Diamond","score":85.5,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/Qwen/Qwen3.5-27B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5-27B","configuration":{"protocol":"Qwen3.5 author-reported model-card evaluation; consult linked evaluation settings"},"artifactId":"76e97b66c81294a6","evidenceId":"qwen35-27b:benchmark:1","confidence":"high","higherIsBetter":true,"date":"2026-04-28T00:00:00.000Z","observedAt":"2026-09-12T16:17:39.480Z"},{"source":"alibaba","name":"SWE-bench Verified","score":72.4,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"MathVision","score":86,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"TAU2-Bench","score":79,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"IFEval","score":91.5,"date":"2026-04-28T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":85.7576,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5 27B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"169e47f5-3d4d-4ad4-8f8b-ab46f0c73f67","evaluatorModelSlug":"qwen3-5-27b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"qwen35-27b:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1403,"arenaVotes":13687,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6ea","name":"Qwen3.5-35B-A3B","slug":"qwen35-35b-a3b","apiIdentifier":"qwen/qwen3.5-35b-a3b","family":"Qwen","version":null,"description":"Qwen3.5-35B-A3B is a multimodal llm from Alibaba. Supports up to 262,144 token context window. Achieves 87.5% on MMLU. Available from $0.16/M input tokens.","organizationSlug":"alibaba","organizationName":"Alibaba","contextWindow":262144,"maxOutputTokens":65536,"inputPricePerMillion":0.1625,"outputPricePerMillion":1.3,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-02-25T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[{"source":"alibaba","name":"MMLU-Pro","score":85.3,"canonicalName":"MMLU-Pro","rawName":"MMLU-Pro","sourceUrl":"https://huggingface.co/Qwen/Qwen3.5-35B-A3B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5-35B-A3B","configuration":{"protocol":"Qwen3.5 author-reported model-card evaluation; consult linked evaluation settings"},"artifactId":"e1dfb292d55fb73a","evidenceId":"qwen35-35b-a3b:benchmark:0","confidence":"high","higherIsBetter":true,"date":"2026-04-28T00:00:00.000Z","observedAt":"2026-09-12T16:17:35.123Z"},{"source":"alibaba","name":"GPQA Diamond","score":84.2,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/Qwen/Qwen3.5-35B-A3B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5-35B-A3B","configuration":{"protocol":"Qwen3.5 author-reported model-card evaluation; consult linked evaluation settings"},"artifactId":"e1dfb292d55fb73a","evidenceId":"qwen35-35b-a3b:benchmark:1","confidence":"high","higherIsBetter":true,"date":"2026-04-28T00:00:00.000Z","observedAt":"2026-09-12T16:17:35.123Z"},{"source":"alibaba","name":"SWE-bench Verified","score":69.2,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"LiveCodeBench","score":74.6,"date":"2026-04-28T00:00:00.000Z"},{"source":"alibaba","name":"MMMU","score":81.4,"canonicalName":"MMMU","rawName":"MMMU","sourceUrl":"https://huggingface.co/Qwen/Qwen3.5-35B-A3B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5-35B-A3B","configuration":{"protocol":"Qwen3.5 author-reported model-card evaluation; consult linked evaluation settings"},"artifactId":"e1dfb292d55fb73a","evidenceId":"qwen35-35b-a3b:benchmark:4","confidence":"high","higherIsBetter":true,"date":"2026-04-28T00:00:00.000Z","observedAt":"2026-09-12T16:17:35.123Z"},{"source":"alibaba","name":"TAU2-Bench","score":81.2,"date":"2026-04-28T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":84.5455,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Qwen3.5 35B A3B (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"46d8315e-1630-463f-ab62-84185fa0faab","evaluatorModelSlug":"qwen3-5-35b-a3b","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"qwen35-35b-a3b:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1396,"arenaVotes":14275,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6c4","name":"LFM2-24B-A2B","slug":"lfm2-24b-a2b","apiIdentifier":"liquid/lfm-2-24b-a2b","family":"LFM2-24B-A2B","version":null,"description":"LFM2-24B-A2B is a large language model from Liquid AI. Supports up to 32,768 token context window. Available from $0.03/M input tokens.","organizationSlug":"liquid-ai","organizationName":"Liquid AI","contextWindow":32768,"maxOutputTokens":null,"inputPricePerMillion":0.03,"outputPricePerMillion":0.12,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-02-25T00:00:00.000Z","tags":["budget-friendly"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":47.3737,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"LFM2 24B A2B","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"fbc58677-e324-4b45-a979-7fd8eec555cd","evaluatorModelSlug":"lfm2-24b-a2b","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"lfm2-24b-a2b:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6ed","name":"Claude Sonnet 4.6","slug":"claude-sonnet-46","apiIdentifier":"anthropic/claude-sonnet-4.6","family":"Claude","version":"4.6","description":"Claude Sonnet 4.6 is a multimodal llm from Anthropic. Supports up to 1,000,000 token context window. Achieves 88.0% on MMLU. Available from $3.00/M input tokens.","organizationSlug":"anthropic","organizationName":"Anthropic","contextWindow":1000000,"maxOutputTokens":128000,"inputPricePerMillion":3,"outputPricePerMillion":15,"capabilities":["text","vision","code","tool-use"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-02-17T00:00:00.000Z","tags":["multimodal","long-context"],"benchmarks":[{"source":"anthropic","name":"MMLU","score":86.7,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"GPQA Diamond","score":89.9,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"SWE-bench Verified","score":72.5,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"Terminal-Bench 2.0","score":51,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"MATH 500","score":54.2,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"LiveCodeBench","score":41.5,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"Berkeley Function Calling","score":88.3,"date":"2026-02-05T00:00:00.000Z"},{"source":"openrouter","name":"HLE","score":30,"date":"2026-02-05T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA-AA Elo","score":1674,"date":"2026-02-05T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":79.899,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Sonnet 4.6 (Non-reasoning, High Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"2e40e695-3cec-43da-83f9-615af30b8e91","evaluatorModelSlug":"claude-sonnet-4-6","reasoning":false,"effort":"high","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"claude-sonnet-46:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1463,"arenaVotes":11255,"aaIntelligenceIndex":52},{"_id":"69e45e9faccbb529d120a6f3","name":"MiniMax M2.5","slug":"minimax-m25-minimax","apiIdentifier":"minimax/minimax-m2.5","family":"MiniMax","version":"M2.5","description":"MiniMax M2.5 is a large language model from MiniMax. Supports up to 196,608 token context window. Available from $0.12/M input tokens.","organizationSlug":"minimax","organizationName":"MiniMax","contextWindow":196608,"maxOutputTokens":65536,"inputPricePerMillion":0.118,"outputPricePerMillion":0.99,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-02-12T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":84.8485,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"MiniMax-M2.5","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"12adec16-19fe-4d92-aeff-5ef3eb7e780a","evaluatorModelSlug":"minimax-m2-5","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"minimax-m25-minimax:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1403,"arenaVotes":19156,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6d9","name":"MiniMax M2.5","slug":"minimax-m25","apiIdentifier":"minimax/minimax-m2.5:free","family":"MiniMax","version":"M2.5","description":"MiniMax M2.5 is a large language model from MiniMax. Supports up to 196,608 token context window. Available free with rate limits on OpenRouter.","organizationSlug":"minimax","organizationName":"MiniMax","contextWindow":196608,"maxOutputTokens":8192,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-02-12T00:00:00.000Z","tags":["long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":84.8485,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"MiniMax-M2.5","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"12adec16-19fe-4d92-aeff-5ef3eb7e780a","evaluatorModelSlug":"minimax-m2-5","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"minimax-m25:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1403,"arenaVotes":19156,"aaIntelligenceIndex":50},{"_id":"69e45e9faccbb529d120a705","name":"GLM 5","slug":"glm-5","apiIdentifier":"z-ai/glm-5","family":"GLM","version":"5","description":"GLM-5 is Z.ai's open-weight language model for agentic engineering, with 744B total and 40B active parameters. Its released checkpoint configuration supports 202,752 token positions; hosted service limits and prices depend on the provider.","organizationSlug":"z-ai","organizationName":"Z.ai","contextWindow":202752,"maxOutputTokens":131072,"inputPricePerMillion":0.72,"outputPricePerMillion":2.3,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-02-11T00:00:00.000Z","tags":["budget-friendly"],"benchmarks":[{"source":"https://huggingface.co/zai-org/GLM-5","name":"GPQA Diamond","score":86,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/zai-org/GLM-5","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"GLM-5","configuration":{"protocol":"GLM-5 author-reported evaluation"},"artifactId":"551ffb1e57ebb8e4","evidenceId":"glm-5:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:39.069Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":82.0202,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"40663ad2-b218-471e-bdd4-a1e0c2360e2b","evaluatorModelSlug":"glm-5","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"glm-5:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1456,"arenaVotes":15839,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a703","name":"Claude Opus 4.6","slug":"claude-opus-46","apiIdentifier":"anthropic/claude-opus-4.6","family":"Claude","version":"4.6","description":"Claude Opus 4.6 is a multimodal LLM from Anthropic with adaptive reasoning capabilities. Achieves 91.3% on GPQA Diamond and 80.8% on SWE-bench Verified. Supports up to 1,000,000 token context window. Available from $5.00/M input tokens.","organizationSlug":"anthropic","organizationName":"Anthropic","contextWindow":1000000,"maxOutputTokens":128000,"inputPricePerMillion":5,"outputPricePerMillion":25,"capabilities":["text","vision","code","tool-use"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-02-04T00:00:00.000Z","tags":["multimodal","long-context"],"benchmarks":[{"source":"anthropic","name":"MMLU","score":89.8,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"GPQA Diamond","score":91.3,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"SWE-bench Verified","score":80.8,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"SWE-bench Pro","score":53.4,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"Terminal-Bench 2.0","score":65.4,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"MATH 500","score":80.7,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"LiveCodeBench","score":55.9,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"Berkeley Function Calling","score":91.9,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"HLE","score":48,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"MCP-Atlas","score":75.8,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"BrowseComp","score":83.7,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"OSWorld-Verified","score":72.7,"date":"2026-02-05T00:00:00.000Z"},{"source":"anthropic","name":"DocVQA","score":91.3,"date":"2026-02-05T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA-AA Elo","score":1619,"date":"2026-02-05T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":84.0404,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Claude Opus 4.6 (Non-reasoning, High Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"4386585e-71b4-4a0c-8a63-afb333419cd6","evaluatorModelSlug":"claude-opus-4-6","reasoning":false,"effort":"high","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"claude-opus-46:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1497,"arenaVotes":19373,"aaIntelligenceIndex":54},{"_id":"69e45e9faccbb529d120a6c9","name":"Step 3.5 Flash","slug":"step-35-flash","apiIdentifier":"stepfun/step-3.5-flash","family":"Step","version":null,"description":"Step 3.5 Flash is a large language model from StepFun. Supports up to 262,144 token context window. Available from $0.10/M input tokens.","organizationSlug":"stepfun","organizationName":"StepFun","contextWindow":262144,"maxOutputTokens":65536,"inputPricePerMillion":0.1,"outputPricePerMillion":0.3,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-01-29T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[],"arenaScore":1392,"arenaVotes":20186,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6da","name":"Trinity Large Preview","slug":"trinity-large-preview","apiIdentifier":"arcee-ai/trinity-large-preview:free","family":"Trinity","version":null,"description":"Trinity Large Preview is a large language model from Arcee AI. Supports up to 131,000 token context window. Available free with rate limits on OpenRouter.","organizationSlug":"arcee-ai","organizationName":"Arcee AI","contextWindow":131000,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-01-27T00:00:00.000Z","tags":["long-context"],"benchmarks":[{"source":"https://huggingface.co/arcee-ai/Trinity-Large-Preview","name":"MMLU","score":87.2,"canonicalName":"MMLU","rawName":"MMLU","sourceUrl":"https://huggingface.co/arcee-ai/Trinity-Large-Preview","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Trinity-Large Preview","configuration":{"protocol":"Trinity-Large Preview author-reported evaluation"},"artifactId":"920d3a0cf9ed4d28","evidenceId":"trinity-large-preview:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:32.009Z"},{"source":"https://huggingface.co/arcee-ai/Trinity-Large-Preview","name":"MMLU-Pro","score":75.2,"canonicalName":"MMLU-Pro","rawName":"MMLU-Pro","sourceUrl":"https://huggingface.co/arcee-ai/Trinity-Large-Preview","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Trinity-Large Preview","configuration":{"protocol":"Trinity-Large Preview author-reported evaluation"},"artifactId":"920d3a0cf9ed4d28","evidenceId":"trinity-large-preview:new-benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:32.009Z"},{"source":"https://huggingface.co/arcee-ai/Trinity-Large-Preview","name":"GPQA Diamond","score":63.3,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/arcee-ai/Trinity-Large-Preview","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Trinity-Large Preview","configuration":{"protocol":"Trinity-Large Preview author-reported evaluation"},"artifactId":"920d3a0cf9ed4d28","evidenceId":"trinity-large-preview:new-benchmark:2","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:32.009Z"}],"arenaScore":1376,"arenaVotes":15015,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6ca","name":"Solar Pro 3","slug":"solar-pro-3","apiIdentifier":"upstage/solar-pro-3","family":"Solar","version":null,"description":"Solar Pro 3 is a large language model from Upstage. Supports up to 128,000 token context window. Available from $0.15/M input tokens.","organizationSlug":"upstage","organizationName":"Upstage","contextWindow":128000,"maxOutputTokens":null,"inputPricePerMillion":0.15,"outputPricePerMillion":0.6,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-01-27T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":7.8155,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Solar Pro 3","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b89c4faf-219e-4171-a1aa-e3bd2fd0a924","evaluatorModelSlug":"solar-pro-3","reasoning":true,"effort":"high","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"solar-pro-3:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":72.4242,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Solar Pro 3","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b89c4faf-219e-4171-a1aa-e3bd2fd0a924","evaluatorModelSlug":"solar-pro-3","reasoning":true,"effort":"high","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"solar-pro-3:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":25.463,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Solar Pro 3","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b89c4faf-219e-4171-a1aa-e3bd2fd0a924","evaluatorModelSlug":"solar-pro-3","reasoning":true,"effort":"high","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"solar-pro-3:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":11.985,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Solar Pro 3","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"b89c4faf-219e-4171-a1aa-e3bd2fd0a924","evaluatorModelSlug":"solar-pro-3","reasoning":true,"effort":"high","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"solar-pro-3:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":26},{"_id":"69e45e9faccbb529d120a6c1","name":"Kimi K2.5","slug":"kimi-k25","apiIdentifier":"moonshotai/kimi-k2.5","family":"Kimi","version":null,"description":"Kimi K2.5 is a multimodal llm from Moonshot AI. Supports up to 262,144 token context window. Available from $0.38/M input tokens.","organizationSlug":"moonshot-ai","organizationName":"Moonshot AI","contextWindow":262144,"maxOutputTokens":65535,"inputPricePerMillion":0.3827,"outputPricePerMillion":1.72,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2026-01-27T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[{"source":"https://huggingface.co/moonshotai/Kimi-K2.5","name":"GPQA Diamond","score":87.6,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2.5","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Kimi K2.5(Thinking)","configuration":{"protocol":"Kimi K2.5(Thinking) author-reported evaluation"},"artifactId":"c7e1a5f4be3bbb49","evidenceId":"kimi-k25:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:22.610Z"},{"source":"https://huggingface.co/moonshotai/Kimi-K2.5","name":"MMLU-Pro","score":87.1,"canonicalName":"MMLU-Pro","rawName":"MMLU-Pro","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2.5","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Kimi K2.5(Thinking)","configuration":{"protocol":"Kimi K2.5(Thinking) author-reported evaluation"},"artifactId":"c7e1a5f4be3bbb49","evidenceId":"kimi-k25:new-benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:22.610Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":87.8788,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Kimi K2.5 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"a550ffca-f89e-4381-ade6-a85dc6a1fb4c","evaluatorModelSlug":"kimi-k2-5","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"kimi-k25:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":45.6929,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Kimi K2.5 (Reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"a550ffca-f89e-4381-ade6-a85dc6a1fb4c","evaluatorModelSlug":"kimi-k2-5","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"kimi-k25:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1451,"arenaVotes":22439,"aaIntelligenceIndex":47},{"_id":"69e45e9faccbb529d120a709","name":"MiniMax M2-her","slug":"minimax-m2-her","apiIdentifier":"minimax/minimax-m2-her","family":"MiniMax","version":"M2-her","description":"MiniMax M2-her is a language-model variant designed for character role-play and interactive storytelling. MiniMax describes work on world consistency, story progression and adaptation to user preferences. It is distinct from the general-purpose MiniMax M2 family.","organizationSlug":"minimax","organizationName":"MiniMax","contextWindow":65536,"maxOutputTokens":2048,"inputPricePerMillion":0.3,"outputPricePerMillion":1.2,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-01-23T00:00:00.000Z","tags":["budget-friendly"],"benchmarks":[],"arenaScore":1346,"arenaVotes":6875,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6cb","name":"Palmyra X5","slug":"palmyra-x5","apiIdentifier":"writer/palmyra-x5","family":"Palmyra","version":null,"description":"Palmyra X5 is Writer's general-purpose AI model with a 1-million-token context window and an 8,192-token output limit. Writer lists direct API pricing of $0.60 input and $6.00 output per million tokens; X5 is scheduled for deprecation on December 14, 2026.","organizationSlug":"writer","organizationName":"Writer","contextWindow":1000000,"maxOutputTokens":8192,"inputPricePerMillion":0.6,"outputPricePerMillion":6,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-01-21T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6fe","name":"LFM2.5-1.2B-Instruct","slug":"lfm25-12b-instruct","apiIdentifier":"liquid/lfm-2.5-1.2b-instruct:free","family":"LFM2.5-1.2B-Instruct","version":null,"description":"LFM2.5-1.2B-Instruct is a large language model from Liquid AI. Supports up to 32,768 token context window. Available free with rate limits on OpenRouter.","organizationSlug":"liquid-ai","organizationName":"Liquid AI","contextWindow":32768,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-01-20T00:00:00.000Z","tags":["instruction-tuned"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":32.6263,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"LFM2.5-1.2B-Instruct","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"48194e0f-8226-4c57-8cb2-2a0fb68a84c9","evaluatorModelSlug":"lfm2-5-1-2b-instruct","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"lfm25-12b-instruct:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6e4","name":"LFM2.5-1.2B-Thinking","slug":"lfm25-12b-thinking","apiIdentifier":"liquid/lfm-2.5-1.2b-thinking:free","family":"LFM2.5-1.2B-Thinking","version":null,"description":"LFM2.5-1.2B-Thinking is a large language model from Liquid AI. Supports up to 32,768 token context window. Available free with rate limits on OpenRouter.","organizationSlug":"liquid-ai","organizationName":"Liquid AI","contextWindow":32768,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","code","extended-thinking"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-01-20T00:00:00.000Z","tags":["reasoning","chain-of-thought"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":33.9394,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"LFM2.5-1.2B-Thinking","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"5a088cde-18e2-4dfa-98dd-d283e1c19654","evaluatorModelSlug":"lfm2-5-1-2b-thinking","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"lfm25-12b-thinking:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a5076319d6168c8158888b2","name":"Pocket TTS","slug":"pocket-tts","apiIdentifier":"kyutai/pocket-tts","family":"Pocket TTS","version":"100M","description":"Pocket TTS is a lightweight text-to-speech model and Python package from Kyutai Labs designed to run efficiently on CPUs with streaming, low-latency audio generation.","organizationSlug":"kyutai","organizationName":"Kyutai","contextWindow":1000000,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text-to-speech","audio","voice-cloning","multilingual","local-inference"],"category":"Audio Model","current":true,"deprecated":false,"releaseDate":"2026-01-13T00:00:00.000Z","tags":["text-to-speech","audio-model","cpu-inference","voice-cloning","open-source","kyutai"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6c6","name":"Olmo 3.1 32B Instruct","slug":"olmo-31-32b-instruct","apiIdentifier":"allenai/olmo-3.1-32b-instruct","family":"Olmo","version":null,"description":"Olmo 3.1 32B Instruct is a large language model from AllenAI. Supports up to 65,536 token context window. Available from $0.20/M input tokens.","organizationSlug":"allenai","organizationName":"AllenAI","contextWindow":65536,"maxOutputTokens":null,"inputPricePerMillion":0.2,"outputPricePerMillion":0.6,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-01-06T00:00:00.000Z","tags":["instruction-tuned","budget-friendly"],"benchmarks":[{"source":"https://huggingface.co/allenai/Olmo-3.1-32B-Instruct","name":"MMLU","score":80.9,"canonicalName":"MMLU","rawName":"MMLU","sourceUrl":"https://huggingface.co/allenai/Olmo-3.1-32B-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Olmo 3.1 32B Instruct","configuration":{"protocol":"Olmo 3.1 32B Instruct author-reported evaluation"},"artifactId":"cdc3004bece03e6b","evidenceId":"olmo-31-32b-instruct:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:25.113Z"},{"source":"https://huggingface.co/allenai/Olmo-3.1-32B-Instruct","name":"GPQA","score":48.6,"canonicalName":"GPQA","rawName":"GPQA","sourceUrl":"https://huggingface.co/allenai/Olmo-3.1-32B-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Olmo 3.1 32B Instruct","configuration":{"protocol":"Olmo 3.1 32B Instruct author-reported evaluation"},"artifactId":"cdc3004bece03e6b","evidenceId":"olmo-31-32b-instruct:new-benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:25.113Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":53.9394,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Olmo 3.1 32B Instruct","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"94a6d26e-a903-47f3-8323-ae422d237bb9","evaluatorModelSlug":"olmo-3-1-32b-instruct","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"olmo-31-32b-instruct:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1331,"arenaVotes":12244,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6fa","name":"Seed 1.6 Flash","slug":"seed-16-flash","apiIdentifier":"bytedance-seed/seed-1.6-flash","family":"Seed","version":null,"description":"Seed 1.6 Flash is a multimodal llm from ByteDance. Supports up to 262,144 token context window. Available from $0.07/M input tokens.","organizationSlug":"bytedance","organizationName":"ByteDance","contextWindow":262144,"maxOutputTokens":32768,"inputPricePerMillion":0.075,"outputPricePerMillion":0.3,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-12-23T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a710","name":"Seed 1.6","slug":"seed-16","apiIdentifier":"bytedance-seed/seed-1.6","family":"Seed","version":null,"description":"Seed 1.6 is a multimodal llm from ByteDance. Supports up to 262,144 token context window. Available from $0.25/M input tokens.","organizationSlug":"bytedance","organizationName":"ByteDance","contextWindow":262144,"maxOutputTokens":32768,"inputPricePerMillion":0.25,"outputPricePerMillion":2,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-12-23T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6f2","name":"Nemotron 3 Nano 30B A3B","slug":"nemotron-3-nano-30b-a3b","apiIdentifier":"nvidia/nemotron-3-nano-30b-a3b:free","family":"Nemotron","version":null,"description":"Nemotron 3 Nano 30B A3B is a large language model from NVIDIA. Supports up to 256,000 token context window. Available free with rate limits on OpenRouter.","organizationSlug":"nvidia","organizationName":"NVIDIA","contextWindow":256000,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-12-14T00:00:00.000Z","tags":["long-context"],"benchmarks":[],"arenaScore":1317,"arenaVotes":15543,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a708","name":"Nemotron 3 Nano 30B A3B","slug":"nemotron-3-nano-30b-a3b-nvidia","apiIdentifier":"nvidia/nemotron-3-nano-30b-a3b","family":"Nemotron","version":null,"description":"Nemotron 3 Nano 30B A3B is a large language model from NVIDIA. Supports up to 262,144 token context window. Available from $0.05/M input tokens.","organizationSlug":"nvidia","organizationName":"NVIDIA","contextWindow":262144,"maxOutputTokens":null,"inputPricePerMillion":0.05,"outputPricePerMillion":0.2,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-12-14T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[],"arenaScore":1317,"arenaVotes":15543,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a702","name":"Ministral 3 14B 2512","slug":"ministral-3-14b-2512","apiIdentifier":"mistralai/ministral-14b-2512","family":"Ministral","version":null,"description":"Ministral 3 14B 2512 is a multimodal llm from Mistral AI. Supports up to 262,144 token context window. Available from $0.20/M input tokens.","organizationSlug":"mistral","organizationName":"Mistral AI","contextWindow":262144,"maxOutputTokens":null,"inputPricePerMillion":0.2,"outputPricePerMillion":0.2,"capabilities":["text","vision","code","tool-use"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-12-02T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6bc","name":"Nova 2 Lite","slug":"nova-2-lite","apiIdentifier":"amazon/nova-2-lite-v1","family":"Nova","version":"Lite","description":"Nova 2 Lite is a multimodal llm from Amazon. Supports up to 1,000,000 token context window. Available from $0.30/M input tokens. Maximum output is unconfirmed: the official overview states 65,536 tokens, while the request schema caps maxTokens at 5,000.","organizationSlug":"amazon","organizationName":"Amazon","contextWindow":1000000,"maxOutputTokens":null,"inputPricePerMillion":0.3,"outputPricePerMillion":2.5,"capabilities":["text","vision","video","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-12-02T00:00:00.000Z","tags":["multimodal","video","budget-friendly","long-context"],"benchmarks":[],"arenaScore":1338,"arenaVotes":12266,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6b6","name":"DeepSeek V3.2 Speciale","slug":"deepseek-v32-speciale","apiIdentifier":"deepseek/deepseek-v3.2-speciale","family":"DeepSeek","version":"Speciale","description":"DeepSeek V3.2 Speciale is a large language model from DeepSeek. Supports up to 163,840 token context window. Achieves 87.1% on MMLU. Available from $0.40/M input tokens.","organizationSlug":"deepseek","organizationName":"DeepSeek","contextWindow":163840,"maxOutputTokens":163840,"inputPricePerMillion":0.4,"outputPricePerMillion":1.2,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-12-01T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"deepseek","name":"MMLU-Pro","score":85,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"AIME 2025","score":96,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"LiveCodeBench","score":88.7,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"HLE","score":30.6,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"HMMT 2025","score":99,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"SWE-bench Verified","score":73,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"Terminal-Bench 2.0","score":46,"date":"2025-12-26T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":87.0707,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V3.2 Speciale","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"527e943a-adc6-4e69-93af-d1608e1b5fed","evaluatorModelSlug":"deepseek-v3-2-speciale","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"deepseek-v32-speciale:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":96.6667,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V3.2 Speciale","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"527e943a-adc6-4e69-93af-d1608e1b5fed","evaluatorModelSlug":"deepseek-v3-2-speciale","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"deepseek-v32-speciale:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1425,"arenaVotes":9078,"aaIntelligenceIndex":42},{"_id":"69e45e9faccbb529d120a6d6","name":"DeepSeek V3.2","slug":"deepseek-v32","apiIdentifier":"deepseek/deepseek-v3.2","family":"DeepSeek","version":null,"description":"DeepSeek V3.2 is a large language model from DeepSeek. Supports up to 163,840 token context window. Achieves 87.1% on MMLU. Available from $0.26/M input tokens.","organizationSlug":"deepseek","organizationName":"DeepSeek","contextWindow":163840,"maxOutputTokens":163840,"inputPricePerMillion":0.259,"outputPricePerMillion":0.42,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-12-01T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"deepseek","name":"MMLU-Pro","score":85,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"GPQA Diamond","score":82.4,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"AIME 2025","score":93.1,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"LiveCodeBench","score":83.3,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"SWE-bench Verified","score":73.1,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"HLE","score":25.1,"date":"2025-12-26T00:00:00.000Z"},{"source":"deepseek","name":"BrowseComp","score":51.4,"date":"2025-12-26T00:00:00.000Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":75.0505,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V3.2 (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"6d9a176d-feb8-4dac-8872-afe32b31897f","evaluatorModelSlug":"deepseek-v3-2","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"deepseek-v32:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":59,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V3.2 (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"6d9a176d-feb8-4dac-8872-afe32b31897f","evaluatorModelSlug":"deepseek-v3-2","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"deepseek-v32:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1424,"arenaVotes":42815,"aaIntelligenceIndex":42},{"_id":"69e45e9faccbb529d120a6f4","name":"Trinity Mini","slug":"trinity-mini","apiIdentifier":"arcee-ai/trinity-mini","family":"Trinity","version":null,"description":"Trinity Mini is a large language model from Arcee AI. Supports up to 131,072 token context window. Available from $0.04/M input tokens.","organizationSlug":"arcee-ai","organizationName":"Arcee AI","contextWindow":131072,"maxOutputTokens":131072,"inputPricePerMillion":0.045,"outputPricePerMillion":0.15,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-12-01T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6e6","name":"Olmo 3 32B Think","slug":"olmo-3-32b-think","apiIdentifier":"allenai/olmo-3-32b-think","family":"Olmo","version":null,"description":"Olmo 3 32B Think is a large language model from AllenAI. Supports up to 65,536 token context window. Available from $0.15/M input tokens.","organizationSlug":"allenai","organizationName":"AllenAI","contextWindow":65536,"maxOutputTokens":65536,"inputPricePerMillion":0.15,"outputPricePerMillion":0.5,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-11-21T00:00:00.000Z","tags":["budget-friendly"],"benchmarks":[{"source":"https://huggingface.co/allenai/Olmo-3-32B-Think","name":"MMLU","score":85.4,"canonicalName":"MMLU","rawName":"MMLU","sourceUrl":"https://huggingface.co/allenai/Olmo-3-32B-Think","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Olmo 3 Think 32B","configuration":{"protocol":"Olmo 3 Think 32B author-reported evaluation"},"artifactId":"212ffbb8c7b86929","evidenceId":"olmo-3-32b-think:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:33.776Z"},{"source":"https://huggingface.co/allenai/Olmo-3-32B-Think","name":"GPQA","score":58.1,"canonicalName":"GPQA","rawName":"GPQA","sourceUrl":"https://huggingface.co/allenai/Olmo-3-32B-Think","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Olmo 3 Think 32B","configuration":{"protocol":"Olmo 3 Think 32B author-reported evaluation"},"artifactId":"212ffbb8c7b86929","evidenceId":"olmo-3-32b-think:new-benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:33.776Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":61.0101,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Olmo 3 32B Think","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"c8a3fa87-735e-49a9-afb1-270c5e9f53f7","evaluatorModelSlug":"olmo-3-32b-think","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"olmo-3-32b-think:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":73.6667,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Olmo 3 32B Think","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"c8a3fa87-735e-49a9-afb1-270c5e9f53f7","evaluatorModelSlug":"olmo-3-32b-think","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"olmo-3-32b-think:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1305,"arenaVotes":5967,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6e1","name":"Kimi K2 Thinking","slug":"kimi-k2-thinking","apiIdentifier":"moonshotai/kimi-k2-thinking","family":"Kimi","version":null,"description":"Kimi K2 Thinking is a large language model from Moonshot AI. Supports up to 262,144 token context window. Available from $0.60/M input tokens.","organizationSlug":"moonshot-ai","organizationName":"Moonshot AI","contextWindow":262144,"maxOutputTokens":262144,"inputPricePerMillion":0.6,"outputPricePerMillion":2.5,"capabilities":["text","code","extended-thinking"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-11-06T00:00:00.000Z","tags":["reasoning","budget-friendly","long-context","chain-of-thought"],"benchmarks":[{"source":"https://huggingface.co/moonshotai/Kimi-K2-Thinking","name":"GPQA","score":84.5,"canonicalName":"GPQA","rawName":"GPQA","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2-Thinking","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"K2 Thinking","configuration":{"protocol":"K2 Thinking author-reported evaluation","tools":"none"},"artifactId":"b47ca20cc743c320","evidenceId":"kimi-k2-thinking:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:32.854Z"},{"source":"https://huggingface.co/moonshotai/Kimi-K2-Thinking","name":"MMLU-Pro","score":84.6,"canonicalName":"MMLU-Pro","rawName":"MMLU-Pro","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2-Thinking","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"K2 Thinking","configuration":{"protocol":"K2 Thinking author-reported evaluation","tools":"none"},"artifactId":"b47ca20cc743c320","evidenceId":"kimi-k2-thinking:new-benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:32.854Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":83.8384,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Kimi K2 Thinking","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"bddebfd3-0a8d-47f5-b722-bc4c2ca5a5dc","evaluatorModelSlug":"kimi-k2-thinking","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"kimi-k2-thinking:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":94.6667,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Kimi K2 Thinking","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"bddebfd3-0a8d-47f5-b722-bc4c2ca5a5dc","evaluatorModelSlug":"kimi-k2-thinking","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"kimi-k2-thinking:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1430,"arenaVotes":47731,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6dc","name":"Nova Premier 1.0","slug":"nova-premier-10","apiIdentifier":"amazon/nova-premier-v1","family":"Nova","version":"1.0","description":"Nova Premier 1.0 is a multimodal llm from Amazon. Supports up to 1,000,000 token context window. Available from $2.50/M input tokens.","organizationSlug":"amazon","organizationName":"Amazon","contextWindow":1000000,"maxOutputTokens":10000,"inputPricePerMillion":2.5,"outputPricePerMillion":12.5,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-10-31T00:00:00.000Z","tags":["multimodal","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6bd","name":"Sonar Pro Search","slug":"sonar-pro-search","apiIdentifier":"perplexity/sonar-pro-search","family":"Sonar","version":"Pro Search","description":"Sonar Pro Search is a multimodal llm from Perplexity AI. Supports up to 200,000 token context window. Achieves 83.0% on MMLU. Available from $3.00/M input tokens.","organizationSlug":"perplexity","organizationName":"Perplexity AI","contextWindow":200000,"maxOutputTokens":8000,"inputPricePerMillion":3,"outputPricePerMillion":15,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-10-30T00:00:00.000Z","tags":["multimodal","long-context"],"benchmarks":[{"source":"perplexity","name":"GPQA Diamond","score":47.1,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"MMLU Pro","score":68.9,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"MATH-500","score":81.7,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"AIME","score":48.7,"date":"2024-11-01T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6cc","name":"Granite 4.0 Micro","slug":"granite-40-micro","apiIdentifier":"ibm-granite/granite-4.0-h-micro","family":"Granite","version":null,"description":"Granite 4.0 Micro is a large language model from IBM. Supports up to 131,000 token context window. Available from $0.02/M input tokens.","organizationSlug":"ibm","organizationName":"IBM","contextWindow":131000,"maxOutputTokens":null,"inputPricePerMillion":0.017,"outputPricePerMillion":0.11,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-10-20T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":33.6364,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Granite 4.0 Micro","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"1fc32894-1060-493b-af94-62bb1068555e","evaluatorModelSlug":"granite-4-0-micro","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"granite-40-micro:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":6,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Granite 4.0 Micro","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"1fc32894-1060-493b-af94-62bb1068555e","evaluatorModelSlug":"granite-4-0-micro","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"granite-40-micro:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6be","name":"ERNIE 4.5 21B A3B Thinking","slug":"ernie-45-21b-a3b-thinking","apiIdentifier":"baidu/ernie-4.5-21b-a3b-thinking","family":"ERNIE","version":null,"description":"ERNIE 4.5 21B A3B Thinking is a large language model from Baidu. Supports up to 131,072 token context window. Available from $0.07/M input tokens.","organizationSlug":"baidu","organizationName":"Baidu","contextWindow":131072,"maxOutputTokens":65536,"inputPricePerMillion":0.07,"outputPricePerMillion":0.28,"capabilities":["text","code","extended-thinking"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-10-09T00:00:00.000Z","tags":["reasoning","budget-friendly","long-context","chain-of-thought"],"benchmarks":[],"arenaScore":1450,"arenaVotes":24211,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6f0","name":"DeepSeek V3.2 Exp","slug":"deepseek-v32-exp","apiIdentifier":"deepseek/deepseek-v3.2-exp","family":"DeepSeek","version":"Exp","description":"DeepSeek V3.2 Exp is a large language model from DeepSeek. Supports up to 163,840 token context window. Achieves 87.1% on MMLU. Available from $0.27/M input tokens.","organizationSlug":"deepseek","organizationName":"DeepSeek","contextWindow":163840,"maxOutputTokens":65536,"inputPricePerMillion":0.27,"outputPricePerMillion":0.41,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-09-29T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"deepseek","name":"MMLU-Pro","score":85,"canonicalName":"MMLU-Pro","rawName":"MMLU-Pro","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.2-Exp","configuration":{"mode":"Reasoning without tools","protocol":"DeepSeek release model-card evaluation"},"artifactId":"259510c63a60818e","evidenceId":"deepseek-v32-exp:benchmark:0","confidence":"high","higherIsBetter":true,"date":"2025-12-26T00:00:00.000Z","observedAt":"2026-09-12T16:34:47.028Z"},{"source":"deepseek","name":"GPQA Diamond","score":79.9,"canonicalName":"GPQA Diamond","rawName":"GPQA-Diamond","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.2-Exp","configuration":{"mode":"Reasoning without tools","protocol":"DeepSeek release model-card evaluation"},"artifactId":"259510c63a60818e","evidenceId":"deepseek-v32-exp:benchmark:1","confidence":"high","higherIsBetter":true,"date":"2025-12-26T00:00:00.000Z","observedAt":"2026-09-12T16:34:47.028Z"},{"source":"deepseek","name":"AIME 2025","score":89.3,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.2-Exp","configuration":{"mode":"Reasoning without tools","protocol":"DeepSeek release model-card evaluation"},"artifactId":"259510c63a60818e","evidenceId":"deepseek-v32-exp:benchmark:2","confidence":"high","higherIsBetter":true,"date":"2025-12-26T00:00:00.000Z","observedAt":"2026-09-12T16:34:47.028Z"},{"source":"deepseek","name":"HMMT 2025","score":83.6,"canonicalName":"HMMT 2025","rawName":"HMMT 2025","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.2-Exp","configuration":{"mode":"Reasoning without tools","protocol":"DeepSeek release model-card evaluation"},"artifactId":"259510c63a60818e","evidenceId":"deepseek-v32-exp:benchmark:3","confidence":"high","higherIsBetter":true,"date":"2025-12-26T00:00:00.000Z","observedAt":"2026-09-12T16:34:47.028Z"},{"source":"deepseek","name":"SWE-bench Verified","score":67.8,"canonicalName":"SWE-bench Verified","rawName":"SWE Verified","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.2-Exp","configuration":{"mode":"Agentic tool use","protocol":"DeepSeek release model-card evaluation"},"artifactId":"259510c63a60818e","evidenceId":"deepseek-v32-exp:benchmark:4","confidence":"high","higherIsBetter":true,"date":"2025-12-26T00:00:00.000Z","observedAt":"2026-09-12T16:34:47.028Z"},{"source":"deepseek","name":"Terminal-Bench","score":37.7,"canonicalName":"Terminal-Bench","rawName":"Terminal-bench","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.2-Exp","configuration":{"mode":"Agentic tool use","protocol":"DeepSeek release model-card evaluation"},"artifactId":"259510c63a60818e","evidenceId":"deepseek-v32-exp:benchmark:5","confidence":"high","higherIsBetter":true,"date":"2025-12-26T00:00:00.000Z","observedAt":"2026-09-12T16:34:47.028Z"},{"source":"deepseek","name":"BrowseComp","score":40.1,"canonicalName":"BrowseComp","rawName":"BrowseComp","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.2-Exp","configuration":{"mode":"Agentic tool use","protocol":"DeepSeek release model-card evaluation"},"artifactId":"259510c63a60818e","evidenceId":"deepseek-v32-exp:benchmark:6","confidence":"high","higherIsBetter":true,"date":"2025-12-26T00:00:00.000Z","observedAt":"2026-09-12T16:34:47.028Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":73.8384,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V3.2 Exp (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"07c35e00-2b12-44c8-91cc-408629cd569e","evaluatorModelSlug":"deepseek-v3-2-0925","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"deepseek-v32-exp:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":57.6667,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V3.2 Exp (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"07c35e00-2b12-44c8-91cc-408629cd569e","evaluatorModelSlug":"deepseek-v3-2-0925","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"deepseek-v32-exp:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1423,"arenaVotes":11952,"aaIntelligenceIndex":42},{"_id":"69e45e9faccbb529d120a706","name":"DeepSeek V3.1 Terminus","slug":"deepseek-v31-terminus","apiIdentifier":"deepseek/deepseek-v3.1-terminus","family":"DeepSeek","version":"Terminus","description":"DeepSeek V3.1 Terminus is a large language model from DeepSeek. Supports up to 163,840 token context window. Achieves 87.1% on MMLU. Available from $0.21/M input tokens.","organizationSlug":"deepseek","organizationName":"DeepSeek","contextWindow":163840,"maxOutputTokens":null,"inputPricePerMillion":0.21,"outputPricePerMillion":0.79,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-09-22T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"deepseek","name":"MMLU-Pro","score":85,"canonicalName":"MMLU-Pro","rawName":"MMLU-Pro","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.1-Terminus","configuration":{"mode":"Reasoning without tools","protocol":"DeepSeek release model-card evaluation"},"artifactId":"095dbc2108ddc63a","evidenceId":"deepseek-v31-terminus:benchmark:0","confidence":"high","higherIsBetter":true,"date":"2025-09-22T00:00:00.000Z","observedAt":"2026-09-12T16:34:46.372Z"},{"source":"deepseek","name":"GPQA Diamond","score":80.7,"canonicalName":"GPQA Diamond","rawName":"GPQA-Diamond","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.1-Terminus","configuration":{"mode":"Reasoning without tools","protocol":"DeepSeek release model-card evaluation"},"artifactId":"095dbc2108ddc63a","evidenceId":"deepseek-v31-terminus:benchmark:1","confidence":"high","higherIsBetter":true,"date":"2025-09-22T00:00:00.000Z","observedAt":"2026-09-12T16:34:46.372Z"},{"source":"deepseek","name":"AIME 2025","score":53.7,"date":"2025-09-22T00:00:00.000Z"},{"source":"deepseek","name":"LiveCodeBench","score":74.9,"canonicalName":"LiveCodeBench","rawName":"LiveCodeBench","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.1-Terminus","configuration":{"mode":"Reasoning without tools","protocol":"DeepSeek release model-card evaluation"},"artifactId":"095dbc2108ddc63a","evidenceId":"deepseek-v31-terminus:benchmark:3","confidence":"high","higherIsBetter":true,"date":"2025-09-22T00:00:00.000Z","observedAt":"2026-09-12T16:34:46.372Z"},{"source":"deepseek","name":"SWE-bench Verified","score":68.4,"canonicalName":"SWE-bench Verified","rawName":"SWE Verified","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.1-Terminus","configuration":{"mode":"Agentic tool use","protocol":"DeepSeek release model-card evaluation"},"artifactId":"095dbc2108ddc63a","evidenceId":"deepseek-v31-terminus:benchmark:4","confidence":"high","higherIsBetter":true,"date":"2025-09-22T00:00:00.000Z","observedAt":"2026-09-12T16:34:46.372Z"},{"source":"deepseek","name":"Terminal-Bench Hard","score":31.8,"date":"2025-09-22T00:00:00.000Z"},{"source":"deepseek","name":"HLE","score":21.7,"canonicalName":"HLE","rawName":"Humanity's Last Exam","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"DeepSeek-V3.1-Terminus","configuration":{"mode":"Reasoning without tools","protocol":"DeepSeek release model-card evaluation"},"artifactId":"095dbc2108ddc63a","evidenceId":"deepseek-v31-terminus:benchmark:6","confidence":"high","higherIsBetter":true,"date":"2025-09-22T00:00:00.000Z","observedAt":"2026-09-12T16:34:46.372Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":75.0505,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V3.1 Terminus (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"dfb9292d-bc7c-4425-a260-4256217e709f","evaluatorModelSlug":"deepseek-v3-1-terminus","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"deepseek-v31-terminus:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":53.6667,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V3.1 Terminus (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"dfb9292d-bc7c-4425-a260-4256217e709f","evaluatorModelSlug":"deepseek-v3-1-terminus","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"deepseek-v31-terminus:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1416,"arenaVotes":3713,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6fb","name":"Kimi K2 0905","slug":"kimi-k2-0905","apiIdentifier":"moonshotai/kimi-k2-0905","family":"Kimi","version":null,"description":"Kimi K2 0905 is a large language model from Moonshot AI. Supports up to 262,144 token context window. Available from $0.40/M input tokens.","organizationSlug":"moonshot-ai","organizationName":"Moonshot AI","contextWindow":262144,"maxOutputTokens":262144,"inputPricePerMillion":0.4,"outputPricePerMillion":2,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-09-04T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":76.6667,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Kimi K2 0905","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"66445f84-b2e3-4202-afdc-92ba0f0e5f36","evaluatorModelSlug":"kimi-k2-0905","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"kimi-k2-0905:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":57.3333,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Kimi K2 0905","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"66445f84-b2e3-4202-afdc-92ba0f0e5f36","evaluatorModelSlug":"kimi-k2-0905","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"kimi-k2-0905:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1418,"arenaVotes":11805,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6bb","name":"Hermes 4 70B","slug":"hermes-4-70b","apiIdentifier":"nousresearch/hermes-4-70b","family":"Hermes","version":"4 70B","description":"Hermes 4 70B is a large language model from Nous Research. Supports up to 131,072 token context window. Available from $0.13/M input tokens.","organizationSlug":"nous-research","organizationName":"Nous Research","contextWindow":131072,"maxOutputTokens":null,"inputPricePerMillion":0.13,"outputPricePerMillion":0.4,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-08-26T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6db","name":"Hermes 4 405B","slug":"hermes-4-405b","apiIdentifier":"nousresearch/hermes-4-405b","family":"Hermes","version":"4 405B","description":"Hermes 4 405B is a large language model from Nous Research. Supports up to 131,072 token context window. Available from $1.00/M input tokens.","organizationSlug":"nous-research","organizationName":"Nous Research","contextWindow":131072,"maxOutputTokens":null,"inputPricePerMillion":1,"outputPricePerMillion":3,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-08-26T00:00:00.000Z","tags":["long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6de","name":"ERNIE 4.5 21B A3B","slug":"ernie-45-21b-a3b","apiIdentifier":"baidu/ernie-4.5-21b-a3b","family":"ERNIE","version":null,"description":"ERNIE 4.5 21B A3B is a large language model from Baidu. Supports up to 120,000 token context window. Available from $0.07/M input tokens.","organizationSlug":"baidu","organizationName":"Baidu","contextWindow":120000,"maxOutputTokens":8000,"inputPricePerMillion":0.07,"outputPricePerMillion":0.28,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-08-12T00:00:00.000Z","tags":["budget-friendly"],"benchmarks":[],"arenaScore":1449,"arenaVotes":9776,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6f8","name":"ERNIE 4.5 VL 28B A3B","slug":"ernie-45-vl-28b-a3b","apiIdentifier":"baidu/ernie-4.5-vl-28b-a3b","family":"ERNIE","version":null,"description":"ERNIE 4.5 VL 28B A3B is a multimodal llm from Baidu. Supports up to 30,000 token context window. Available from $0.14/M input tokens.","organizationSlug":"baidu","organizationName":"Baidu","contextWindow":30000,"maxOutputTokens":8000,"inputPricePerMillion":0.14,"outputPricePerMillion":0.56,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-08-12T00:00:00.000Z","tags":["multimodal","budget-friendly"],"benchmarks":[],"arenaScore":1419,"arenaVotes":4730,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6cd","name":"Jamba Large 1.7","slug":"jamba-large-17","apiIdentifier":"ai21/jamba-large-1.7","family":"Jamba","version":null,"description":"Jamba Large 1.7 is a large language model from AI21 Labs. Supports up to 256,000 token context window. Available from $2.00/M input tokens.","organizationSlug":"ai21","organizationName":"AI21 Labs","contextWindow":256000,"maxOutputTokens":4096,"inputPricePerMillion":2,"outputPricePerMillion":8,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-08-08T00:00:00.000Z","tags":["long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69fd6d1928ad1127d4ac8a66","name":"Kronos","slug":"kronos","apiIdentifier":"shiyu-coder/kronos","family":"Kronos","version":"2025","description":"Kronos is a foundation model for financial K-line market sequences. It tokenizes continuous market records and uses autoregressive pre-training for forecasting, volatility, and synthetic time-series tasks.","organizationSlug":"tsinghua-university","organizationName":"Tsinghua University","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["time-series-forecasting","financial-modeling","synthetic-data","volatility-prediction"],"category":"Time Series Foundation Model","current":true,"deprecated":false,"releaseDate":"2025-08-02T00:00:00.000Z","tags":["financial-ai","time-series","foundation-model","forecasting","open-source"],"benchmarks":[{"source":"official-arxiv","name":"RankIC improvement over leading TSFM","score":93,"date":"2025-08-02T00:00:00.000Z"},{"source":"official-arxiv","name":"Volatility MAE reduction","score":9,"date":"2025-08-02T00:00:00.000Z"},{"source":"official-arxiv","name":"Synthetic K-line fidelity improvement","score":22,"date":"2025-08-02T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a711","name":"Kimi K2 0711","slug":"kimi-k2-0711","apiIdentifier":"moonshotai/kimi-k2","family":"Kimi","version":null,"description":"Kimi K2 0711 is a large language model from Moonshot AI. Supports up to 131,072 token context window. Available from $0.57/M input tokens.","organizationSlug":"moonshot-ai","organizationName":"Moonshot AI","contextWindow":131072,"maxOutputTokens":32768,"inputPricePerMillion":0.57,"outputPricePerMillion":2.3,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-07-11T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"https://huggingface.co/moonshotai/Kimi-K2-Instruct","name":"GPQA Diamond","score":75.1,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Kimi K2 Instruct","configuration":{"protocol":"Kimi K2 Instruct author-reported evaluation","metric":"Avg@8"},"artifactId":"80c63c55e5fb028f","evidenceId":"kimi-k2-0711:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:39.699Z"},{"source":"https://huggingface.co/moonshotai/Kimi-K2-Instruct","name":"MMLU","score":89.5,"canonicalName":"MMLU","rawName":"MMLU","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Kimi K2 Instruct","configuration":{"protocol":"Kimi K2 Instruct author-reported evaluation","metric":"EM"},"artifactId":"80c63c55e5fb028f","evidenceId":"kimi-k2-0711:new-benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:39.699Z"},{"source":"https://huggingface.co/moonshotai/Kimi-K2-Instruct","name":"MMLU-Pro","score":81.1,"canonicalName":"MMLU-Pro","rawName":"MMLU-Pro","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Kimi K2 Instruct","configuration":{"protocol":"Kimi K2 Instruct author-reported evaluation","metric":"EM"},"artifactId":"80c63c55e5fb028f","evidenceId":"kimi-k2-0711:new-benchmark:2","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:39.699Z"}],"arenaScore":1417,"arenaVotes":27669,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6ce","name":"Hunyuan A13B Instruct","slug":"hunyuan-a13b-instruct","apiIdentifier":"tencent/hunyuan-a13b-instruct","family":"Hunyuan","version":null,"description":"Hunyuan A13B Instruct is a large language model from Tencent. Supports up to 131,072 token context window. Available from $0.14/M input tokens.","organizationSlug":"tencent","organizationName":"Tencent","contextWindow":131072,"maxOutputTokens":131072,"inputPricePerMillion":0.14,"outputPricePerMillion":0.57,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-07-08T00:00:00.000Z","tags":["instruction-tuned","budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a1e5e6e2ae589a72aa3c738","name":"Cube","slug":"cube","apiIdentifier":"Roblox/cube","family":"Cube","version":"3D v0.5","description":"Cube is Roblox’s foundation-model project for 3D intelligence. It includes text-to-3D generation and part-controllable 3D asset generation for creator workflows.","organizationSlug":"roblox","organizationName":"Roblox","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text-to-3d","3d-generation","part-controllable-generation","shape-generation"],"category":"3D Model","current":true,"deprecated":false,"releaseDate":"2025-07-01T00:00:00.000Z","tags":["text-to-3d","3d-generation","foundation-model","roblox","generative-ai"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a70e","name":"ERNIE 4.5 VL 424B A47B","slug":"ernie-45-vl-424b-a47b","apiIdentifier":"baidu/ernie-4.5-vl-424b-a47b","family":"ERNIE","version":null,"description":"ERNIE 4.5 VL 424B A47B is a multimodal llm from Baidu. Supports up to 123,000 token context window. Available from $0.42/M input tokens.","organizationSlug":"baidu","organizationName":"Baidu","contextWindow":123000,"maxOutputTokens":16000,"inputPricePerMillion":0.42,"outputPricePerMillion":1.25,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-06-30T00:00:00.000Z","tags":["multimodal","budget-friendly"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a70a","name":"Spotlight","slug":"spotlight","apiIdentifier":"arcee-ai/spotlight","family":"Spotlight","version":null,"description":"Spotlight is a multimodal llm from Arcee AI. Supports up to 131,072 token context window. Available from $0.18/M input tokens.","organizationSlug":"arcee-ai","organizationName":"Arcee AI","contextWindow":131072,"maxOutputTokens":65537,"inputPricePerMillion":0.18,"outputPricePerMillion":0.18,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-05-05T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6b4","name":"Llama Guard 4 12B","slug":"llama-guard-4-12b","apiIdentifier":"meta-llama/llama-guard-4-12b","family":"Llama","version":"4 12B","description":"Llama Guard 4 12B is a Meta multimodal safety-classification model for text and image prompt or response moderation using MLCommons hazard categories.","organizationSlug":"meta","organizationName":"Meta","contextWindow":163840,"maxOutputTokens":40,"inputPricePerMillion":0.18,"outputPricePerMillion":0.18,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-04-30T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context","safety","moderation","guardrails"],"benchmarks":[{"source":"meta","name":"LG4 English output-filter F1","score":61,"sourceUrl":"https://huggingface.co/meta-llama/Llama-Guard-4-12B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama Guard 4 12B","configuration":{"task":"output-filter","subset":"English","metric":"F1"},"artifactId":"4cb15753778b29fa","evidenceId":"llama-guard-4-12b:benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:34:45.065Z"},{"source":"meta","name":"LG4 multilingual output-filter F1","score":51,"sourceUrl":"https://huggingface.co/meta-llama/Llama-Guard-4-12B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama Guard 4 12B","configuration":{"task":"output-filter","subset":"Multilingual","metric":"F1"},"artifactId":"4cb15753778b29fa","evidenceId":"llama-guard-4-12b:benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:34:45.065Z"},{"source":"meta","name":"LG4 single-image output-filter F1","score":38,"sourceUrl":"https://huggingface.co/meta-llama/Llama-Guard-4-12B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama Guard 4 12B","configuration":{"task":"output-filter","subset":"Single-image","metric":"F1"},"artifactId":"4cb15753778b29fa","evidenceId":"llama-guard-4-12b:benchmark:2","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:34:45.065Z"},{"source":"meta","name":"LG4 multi-image output-filter F1","score":52,"sourceUrl":"https://huggingface.co/meta-llama/Llama-Guard-4-12B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama Guard 4 12B","configuration":{"task":"output-filter","subset":"Multi-image","metric":"F1"},"artifactId":"4cb15753778b29fa","evidenceId":"llama-guard-4-12b:benchmark:3","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:34:45.065Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e458d996953a3e27dd4059","name":"Qwen3-VL","slug":"qwen3vl","apiIdentifier":"qwen-vl-max","family":"Qwen3","version":"3","description":"Qwen3-VL is Alibaba's multimodal vision-language model from the Qwen3 family. It processes images, videos, and text together, excelling at document understanding, chart reading, OCR, and visual reasoning tasks across multiple languages.","organizationSlug":"alibaba","organizationName":"Alibaba","contextWindow":131072,"maxOutputTokens":8192,"inputPricePerMillion":0.2,"outputPricePerMillion":0.6,"capabilities":["text","vision","code","tool-use"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-04-28T00:00:00.000Z","tags":["vision-language-model","multimodal","ocr","document-understanding","video-understanding","chinese-ai","alibaba","qwen","open-source","visual-reasoning"],"benchmarks":[{"source":"openrouter","name":"MMMU","score":70.3,"date":"2025-04-28T00:00:00.000Z"},{"source":"openrouter","name":"DocVQA","score":94.1,"date":"2025-04-28T00:00:00.000Z"},{"source":"openrouter","name":"ChartQA","score":86.5,"date":"2025-04-28T00:00:00.000Z"},{"source":"openrouter","name":"OCRBench","score":88.7,"date":"2025-04-28T00:00:00.000Z"},{"source":"openrouter","name":"MathVista","score":74.8,"date":"2025-04-28T00:00:00.000Z"},{"source":"openrouter","name":"RealWorldQA","score":75.2,"date":"2025-04-28T00:00:00.000Z"},{"source":"openrouter","name":"Video-MME","score":69.8,"date":"2025-04-28T00:00:00.000Z"}],"arenaScore":1416,"arenaVotes":11533,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6d4","name":"Llama 4 Maverick","slug":"llama-4-maverick","apiIdentifier":"meta-llama/llama-4-maverick","family":"Llama","version":"Maverick","description":"Llama 4 Maverick is a multimodal llm from Meta. Supports up to 1,048,576 token context window. Achieves 88.4% on MMLU. Available from $0.15/M input tokens.","organizationSlug":"meta","organizationName":"Meta","contextWindow":1048576,"maxOutputTokens":16384,"inputPricePerMillion":0.15,"outputPricePerMillion":0.6,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-04-05T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[{"source":"meta","name":"MMLU","score":85.5,"sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Maverick Base","configuration":{"checkpoint":"base","notListedVariant":true},"artifactId":"65471a2c7ad581fb","evidenceId":"llama-4-maverick:benchmark:0","confidence":"high","comparisonEligible":false,"exclusionReason":"This observation belongs to the base checkpoint, not the listed instruction-tuned model","date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.118Z"},{"source":"meta","name":"MMLU Pro","score":80.5,"canonicalName":"MMLU Pro","rawName":"MMLU Pro","sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Maverick Instruct","configuration":{"shots":0,"metric":"macro_avg/acc","checkpoint":"instruction-tuned"},"artifactId":"65471a2c7ad581fb","evidenceId":"llama-4-maverick:benchmark:1","confidence":"high","higherIsBetter":true,"date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.118Z"},{"source":"meta","name":"GPQA Diamond","score":69.8,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Maverick Instruct","configuration":{"shots":0,"metric":"accuracy","checkpoint":"instruction-tuned"},"artifactId":"65471a2c7ad581fb","evidenceId":"llama-4-maverick:benchmark:2","confidence":"high","higherIsBetter":true,"date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.118Z"},{"source":"meta","name":"MATH","score":61.2,"sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Maverick Base","configuration":{"checkpoint":"base","notListedVariant":true},"artifactId":"65471a2c7ad581fb","evidenceId":"llama-4-maverick:benchmark:3","confidence":"high","comparisonEligible":false,"exclusionReason":"This observation belongs to the base checkpoint, not the listed instruction-tuned model","date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.118Z"},{"source":"meta","name":"LiveCodeBench (10/01/2024-02/01/2025)","score":43.4,"canonicalName":"LiveCodeBench (10/01/2024-02/01/2025)","rawName":"LiveCodeBench (10/01/2024-02/01/2025)","sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Maverick Instruct","configuration":{"shots":0,"metric":"pass@1","checkpoint":"instruction-tuned"},"benchmarkVersion":"2024-10-01 to 2025-02-01","artifactId":"65471a2c7ad581fb","evidenceId":"llama-4-maverick:benchmark:4","confidence":"high","higherIsBetter":true,"date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.118Z"},{"source":"meta","name":"MMMU","score":73.4,"canonicalName":"MMMU","rawName":"MMMU","sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Maverick Instruct","configuration":{"shots":0,"metric":"accuracy","checkpoint":"instruction-tuned"},"artifactId":"65471a2c7ad581fb","evidenceId":"llama-4-maverick:benchmark:5","confidence":"high","higherIsBetter":true,"date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.118Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":9.299,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Llama 4 Maverick","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"922c69c7-9037-43c6-8bcf-a1c555e7f3eb","evaluatorModelSlug":"llama-4-maverick","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"llama-4-maverick:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":67.0707,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Maverick","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"922c69c7-9037-43c6-8bcf-a1c555e7f3eb","evaluatorModelSlug":"llama-4-maverick","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"llama-4-maverick:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":31.713,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Maverick","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"922c69c7-9037-43c6-8bcf-a1c555e7f3eb","evaluatorModelSlug":"llama-4-maverick","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"llama-4-maverick:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":7.8652,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Maverick","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"922c69c7-9037-43c6-8bcf-a1c555e7f3eb","evaluatorModelSlug":"llama-4-maverick","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"llama-4-maverick:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":19.3333,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Maverick","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"922c69c7-9037-43c6-8bcf-a1c555e7f3eb","evaluatorModelSlug":"llama-4-maverick","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"llama-4-maverick:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1327,"arenaVotes":40010,"aaIntelligenceIndex":18},{"_id":"69e45e9faccbb529d120a6ee","name":"Llama 4 Scout","slug":"llama-4-scout","apiIdentifier":"meta-llama/llama-4-scout","family":"Llama","version":"Scout","description":"Llama 4 Scout is a multimodal llm from Meta. Supports up to 327,680 token context window. Achieves 82.6% on MMLU. Available from $0.08/M input tokens.","organizationSlug":"meta","organizationName":"Meta","contextWindow":327680,"maxOutputTokens":16384,"inputPricePerMillion":0.08,"outputPricePerMillion":0.3,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-04-05T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[{"source":"meta","name":"MMLU","score":79.6,"sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout Base","configuration":{"checkpoint":"base","notListedVariant":true},"artifactId":"8cb15ff6ddd4d6a6","evidenceId":"llama-4-scout:benchmark:0","confidence":"high","comparisonEligible":false,"exclusionReason":"This observation belongs to the base checkpoint, not the listed instruction-tuned model","date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.277Z"},{"source":"meta","name":"MMLU Pro","score":74.3,"canonicalName":"MMLU Pro","rawName":"MMLU Pro","sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout Instruct","configuration":{"shots":0,"metric":"macro_avg/acc","checkpoint":"instruction-tuned"},"artifactId":"8cb15ff6ddd4d6a6","evidenceId":"llama-4-scout:benchmark:1","confidence":"high","higherIsBetter":true,"date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.277Z"},{"source":"meta","name":"GPQA Diamond","score":57.2,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout Instruct","configuration":{"shots":0,"metric":"accuracy","checkpoint":"instruction-tuned"},"artifactId":"8cb15ff6ddd4d6a6","evidenceId":"llama-4-scout:benchmark:2","confidence":"high","higherIsBetter":true,"date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.277Z"},{"source":"meta","name":"MATH","score":50.3,"sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout Base","configuration":{"checkpoint":"base","notListedVariant":true},"artifactId":"8cb15ff6ddd4d6a6","evidenceId":"llama-4-scout:benchmark:3","confidence":"high","comparisonEligible":false,"exclusionReason":"This observation belongs to the base checkpoint, not the listed instruction-tuned model","date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.277Z"},{"source":"meta","name":"LiveCodeBench (10/01/2024-02/01/2025)","score":32.8,"canonicalName":"LiveCodeBench (10/01/2024-02/01/2025)","rawName":"LiveCodeBench (10/01/2024-02/01/2025)","sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout Instruct","configuration":{"shots":0,"metric":"pass@1","checkpoint":"instruction-tuned"},"benchmarkVersion":"2024-10-01 to 2025-02-01","artifactId":"8cb15ff6ddd4d6a6","evidenceId":"llama-4-scout:benchmark:4","confidence":"high","higherIsBetter":true,"date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.277Z"},{"source":"meta","name":"MMMU","score":69.4,"canonicalName":"MMMU","rawName":"MMMU","sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout Instruct","configuration":{"shots":0,"metric":"accuracy","checkpoint":"instruction-tuned"},"artifactId":"8cb15ff6ddd4d6a6","evidenceId":"llama-4-scout:benchmark:5","confidence":"high","higherIsBetter":true,"date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.277Z"},{"source":"meta","name":"MGSM","score":90.6,"canonicalName":"MGSM","rawName":"MGSM","sourceUrl":"https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout Instruct","configuration":{"shots":0,"metric":"average/em","checkpoint":"instruction-tuned"},"artifactId":"8cb15ff6ddd4d6a6","evidenceId":"llama-4-scout:benchmark:6","confidence":"high","higherIsBetter":true,"date":"2025-04-05T00:00:00.000Z","observedAt":"2026-09-12T16:34:45.277Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":6.4513,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Llama 4 Scout","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"adf9a85e-abc3-4f28-937b-db6655cc5238","evaluatorModelSlug":"llama-4-scout","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"llama-4-scout:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":58.6869,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"adf9a85e-abc3-4f28-937b-db6655cc5238","evaluatorModelSlug":"llama-4-scout","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"llama-4-scout:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":21.2963,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"adf9a85e-abc3-4f28-937b-db6655cc5238","evaluatorModelSlug":"llama-4-scout","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"llama-4-scout:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":3.7453,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"adf9a85e-abc3-4f28-937b-db6655cc5238","evaluatorModelSlug":"llama-4-scout","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"llama-4-scout:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":14,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Llama 4 Scout","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"adf9a85e-abc3-4f28-937b-db6655cc5238","evaluatorModelSlug":"llama-4-scout","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"llama-4-scout:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1322,"arenaVotes":30316,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6c2","name":"Command A","slug":"command-a","apiIdentifier":"cohere/command-a","family":"Command","version":null,"description":"Command A is a large language model from Cohere. Supports up to 256,000 token context window. Achieves 80.1% on MMLU. Available from $2.50/M input tokens.","organizationSlug":"cohere","organizationName":"Cohere","contextWindow":256000,"maxOutputTokens":8192,"inputPricePerMillion":2.5,"outputPricePerMillion":10,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-03-13T00:00:00.000Z","tags":["long-context"],"benchmarks":[{"source":"openrouter","name":"MMLU","score":80.1,"date":"2025-03-13T00:00:00.000Z"}],"arenaScore":1354,"arenaVotes":56351,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6e5","name":"Reka Flash 3","slug":"reka-flash-3","apiIdentifier":"rekaai/reka-flash-3","family":"Reka","version":null,"description":"Reka Flash 3 is a large language model from Reka. Supports up to 65,536 token context window. Available from $0.10/M input tokens.","organizationSlug":"reka","organizationName":"Reka","contextWindow":65536,"maxOutputTokens":65536,"inputPricePerMillion":0.1,"outputPricePerMillion":0.2,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-03-12T00:00:00.000Z","tags":["budget-friendly"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":52.9293,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Reka Flash 3","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"dbbcf240-a69e-4078-8f16-7c94c7a8c514","evaluatorModelSlug":"reka-flash-3","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"reka-flash-3:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":33.6667,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Reka Flash 3","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"dbbcf240-a69e-4078-8f16-7c94c7a8c514","evaluatorModelSlug":"reka-flash-3","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"reka-flash-3:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6dd","name":"Sonar Reasoning Pro","slug":"sonar-reasoning-pro","apiIdentifier":"perplexity/sonar-reasoning-pro","family":"Sonar","version":"Reasoning Pro","description":"Sonar Reasoning Pro is a multimodal llm from Perplexity AI. Supports up to 128,000 token context window. Achieves 83.0% on MMLU. Available from $2.00/M input tokens.","organizationSlug":"perplexity","organizationName":"Perplexity AI","contextWindow":128000,"maxOutputTokens":null,"inputPricePerMillion":2,"outputPricePerMillion":8,"capabilities":["text","vision","code","extended-thinking"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-03-07T00:00:00.000Z","tags":["reasoning","multimodal","long-context"],"benchmarks":[{"source":"perplexity","name":"GPQA Diamond","score":62.3,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"MATH-500","score":92.1,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"AIME","score":77,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"MMLU Pro","score":79,"date":"2024-11-01T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6f7","name":"Sonar Pro","slug":"sonar-pro","apiIdentifier":"perplexity/sonar-pro","family":"Sonar","version":"Pro","description":"Sonar Pro is a multimodal llm from Perplexity AI. Supports up to 200,000 token context window. Achieves 83.0% on MMLU. Available from $3.00/M input tokens.","organizationSlug":"perplexity","organizationName":"Perplexity AI","contextWindow":200000,"maxOutputTokens":8000,"inputPricePerMillion":3,"outputPricePerMillion":15,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2025-03-07T00:00:00.000Z","tags":["multimodal","long-context"],"benchmarks":[{"source":"perplexity","name":"GPQA Diamond","score":57.8,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"MMLU Pro","score":75.5,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"MATH-500","score":74.5,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"AIME","score":29,"date":"2024-11-01T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a70d","name":"Sonar Deep Research","slug":"sonar-deep-research","apiIdentifier":"perplexity/sonar-deep-research","family":"Sonar","version":"Deep Research","description":"Sonar Deep Research is a large language model from Perplexity AI. Supports up to 128,000 token context window. Achieves 83.0% on MMLU. Available from $2.00/M input tokens.","organizationSlug":"perplexity","organizationName":"Perplexity AI","contextWindow":128000,"maxOutputTokens":null,"inputPricePerMillion":2,"outputPricePerMillion":8,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-03-07T00:00:00.000Z","tags":["long-context"],"benchmarks":[{"source":"perplexity","name":"GPQA Diamond","score":47.1,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"MMLU Pro","score":68.9,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"MATH-500","score":81.7,"date":"2024-11-01T00:00:00.000Z"},{"source":"perplexity","name":"AIME","score":48.7,"date":"2024-11-01T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a704","name":"Llama Guard 3 8B","slug":"llama-guard-3-8b","apiIdentifier":"meta-llama/llama-guard-3-8b","family":"Llama","version":"3 8B","description":"Llama Guard 3 8B is a Meta safety-classification model for prompt and response moderation across English, multilingual, search, and code-interpreter use cases.","organizationSlug":"meta","organizationName":"Meta","contextWindow":131072,"maxOutputTokens":100,"inputPricePerMillion":0.48,"outputPricePerMillion":0.03,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-02-12T00:00:00.000Z","tags":["budget-friendly","long-context","safety","moderation","guardrails"],"benchmarks":[{"source":"meta","name":"LG3 English safety F1","score":93.9,"sourceUrl":"https://huggingface.co/meta-llama/Llama-Guard-3-8B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama Guard 3 8B","configuration":{"metric":"F1","task":"English output safety","sourceScale":"0-1","normalization":"source fraction multiplied by 100"},"artifactId":"7f2f59418bb5d5ca","evidenceId":"llama-guard-3-8b:benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:34:45.640Z"},{"source":"meta","name":"LG3 English safety AUPRC","score":98.5,"sourceUrl":"https://huggingface.co/meta-llama/Llama-Guard-3-8B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama Guard 3 8B","configuration":{"metric":"AUPRC","task":"English output safety","sourceScale":"0-1","normalization":"source fraction multiplied by 100"},"artifactId":"7f2f59418bb5d5ca","evidenceId":"llama-guard-3-8b:benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:34:45.640Z"},{"source":"meta","name":"LG3 tool-use safety F1","score":85.6,"sourceUrl":"https://huggingface.co/meta-llama/Llama-Guard-3-8B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama Guard 3 8B","configuration":{"metric":"F1","task":"Search tool calls","sourceScale":"0-1","normalization":"source fraction multiplied by 100"},"artifactId":"7f2f59418bb5d5ca","evidenceId":"llama-guard-3-8b:benchmark:2","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:34:45.640Z"},{"source":"meta","name":"LG3 code-interpreter safety F1","score":88.5,"sourceUrl":"https://huggingface.co/meta-llama/Llama-Guard-3-8B","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Llama Guard 3 8B","configuration":{"metric":"F1","task":"Code interpreter abuse","sourceScale":"0-1","normalization":"source fraction multiplied by 100"},"artifactId":"7f2f59418bb5d5ca","evidenceId":"llama-guard-3-8b:benchmark:3","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:34:45.640Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6c7","name":"Phi 4","slug":"phi-4","apiIdentifier":"microsoft/phi-4","family":"Phi","version":"4","description":"Phi 4 is a large language model from Microsoft. Supports up to 16,384 token context window. Available from $0.07/M input tokens.","organizationSlug":"microsoft","organizationName":"Microsoft","contextWindow":16384,"maxOutputTokens":16384,"inputPricePerMillion":0.065,"outputPricePerMillion":0.14,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2025-01-10T00:00:00.000Z","tags":["budget-friendly"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":57.4747,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Phi-4","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"ee708f92-374e-4123-b900-e22d7b2afc19","evaluatorModelSlug":"phi-4","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"phi-4:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":18,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Phi-4","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"ee708f92-374e-4123-b900-e22d7b2afc19","evaluatorModelSlug":"phi-4","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"phi-4:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1256,"arenaVotes":24126,"aaIntelligenceIndex":null},{"_id":"69fd6d1928ad1127d4ac8a67","name":"TabPFN","slug":"tabpfn","apiIdentifier":"priorlabs/tabpfn-2.5","family":"TabPFN","version":"2.5","description":"TabPFN is Prior Labs’ tabular foundation model for classification and regression on structured data. It returns predictions in one forward pass and reduces the need for manual model tuning.","organizationSlug":"prior-labs","organizationName":"Prior Labs","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["classification","regression","anomaly-detection","time-series-forecasting","synthetic-data"],"category":"Tabular Foundation Model","current":true,"deprecated":false,"releaseDate":"2025-01-01T00:00:00.000Z","tags":["tabular-ai","foundation-model","classification","regression","structured-data"],"benchmarks":[{"source":"official-prior-labs","name":"TabArena vs tuned tree-based models","score":100,"date":"2026-05-08T04:56:57.493Z"},{"source":"official-prior-labs","name":"Maximum supported data points","score":50000,"date":"2026-05-08T04:56:57.493Z"},{"source":"official-prior-labs","name":"Maximum supported features","score":2000,"date":"2026-05-08T04:56:57.493Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6e2","name":"Command R7B (12-2024)","slug":"command-r7b-12-2024","apiIdentifier":"cohere/command-r7b-12-2024","family":"Command","version":null,"description":"Command R7B (12-2024) is a large language model from Cohere. Supports up to 128,000 token context window. Available from $0.04/M input tokens.","organizationSlug":"cohere","organizationName":"Cohere","contextWindow":128000,"maxOutputTokens":4000,"inputPricePerMillion":0.0375,"outputPricePerMillion":0.15,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2024-12-14T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[],"arenaScore":1226,"arenaVotes":54036,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6f6","name":"Nova Lite 1.0","slug":"nova-lite-10","apiIdentifier":"amazon/nova-lite-v1","family":"Nova","version":"1.0","description":"Nova Lite 1.0 is a multimodal llm from Amazon. Supports up to 300,000 token context window. Achieves 72.0% on MMLU. Available from $0.06/M input tokens.","organizationSlug":"amazon","organizationName":"Amazon","contextWindow":300000,"maxOutputTokens":10000,"inputPricePerMillion":0.06,"outputPricePerMillion":0.24,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":true,"deprecated":false,"releaseDate":"2024-12-05T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[{"source":"openrouter","name":"MMLU","score":72,"date":"2024-12-03T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a70c","name":"Nova Micro 1.0","slug":"nova-micro-10","apiIdentifier":"amazon/nova-micro-v1","family":"Nova","version":"1.0","description":"Nova Micro 1.0 is Amazon's text-only model for fast, low-cost tasks including summarization, translation, and classification. Amazon documents a 128K-token context window for the Bedrock model.","organizationSlug":"amazon","organizationName":"Amazon","contextWindow":128000,"maxOutputTokens":null,"inputPricePerMillion":0.035,"outputPricePerMillion":0.14,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2024-12-05T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"openrouter","name":"MMLU","score":69.5,"date":"2024-12-03T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6c8","name":"Inflection 3 Pi","slug":"inflection-3-pi","apiIdentifier":"inflection/inflection-3-pi","family":"Inflection","version":null,"description":"Inflection 3 Pi is a large language model from Inflection AI. Supports up to 8,000 token context window. Available from $2.50/M input tokens.","organizationSlug":"inflection","organizationName":"Inflection AI","contextWindow":8000,"maxOutputTokens":1024,"inputPricePerMillion":2.5,"outputPricePerMillion":10,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2024-10-11T00:00:00.000Z","tags":["general-purpose"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6e8","name":"Inflection 3 Productivity","slug":"inflection-3-productivity","apiIdentifier":"inflection/inflection-3-productivity","family":"Inflection","version":null,"description":"Inflection 3 Productivity is a large language model from Inflection AI. Supports up to 8,000 token context window. Available from $2.50/M input tokens.","organizationSlug":"inflection","organizationName":"Inflection AI","contextWindow":8000,"maxOutputTokens":1024,"inputPricePerMillion":2.5,"outputPricePerMillion":10,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2024-10-11T00:00:00.000Z","tags":["general-purpose"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6fc","name":"Command R+ (08-2024)","slug":"command-r-08-2024","apiIdentifier":"cohere/command-r-plus-08-2024","family":"Command","version":null,"description":"Command R+ (08-2024) is a large language model from Cohere. Supports up to 128,000 token context window. Achieves 76.3% on MMLU. Available from $2.50/M input tokens.","organizationSlug":"cohere","organizationName":"Cohere","contextWindow":128000,"maxOutputTokens":4000,"inputPricePerMillion":2.5,"outputPricePerMillion":10,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2024-08-30T00:00:00.000Z","tags":["long-context"],"benchmarks":[{"source":"openrouter","name":"MMLU","score":76.3,"date":"2024-06-13T00:00:00.000Z"}],"arenaScore":1249,"arenaVotes":10140,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a712","name":"Command R (08-2024)","slug":"command-r-08-2024-cohere","apiIdentifier":"cohere/command-r-08-2024","family":"Command","version":null,"description":"Command R (08-2024) is a large language model from Cohere. Supports up to 128,000 token context window. Achieves 68.4% on MMLU. Available from $0.15/M input tokens.","organizationSlug":"cohere","organizationName":"Cohere","contextWindow":128000,"maxOutputTokens":4000,"inputPricePerMillion":0.15,"outputPricePerMillion":0.6,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2024-08-30T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"openrouter","name":"MMLU","score":68.4,"date":"2024-06-13T00:00:00.000Z"}],"arenaScore":1249,"arenaVotes":10140,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6f5","name":"Hermes 3 70B Instruct","slug":"hermes-3-70b-instruct","apiIdentifier":"nousresearch/hermes-3-llama-3.1-70b","family":"Hermes","version":"3 70B Instruct","description":"Hermes 3 70B Instruct is a large language model from Nous Research. Supports up to 131,072 token context window. Achieves 82.0% on MMLU. Available from $0.30/M input tokens.","organizationSlug":"nous-research","organizationName":"Nous Research","contextWindow":131072,"maxOutputTokens":null,"inputPricePerMillion":0.3,"outputPricePerMillion":0.3,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2024-08-18T00:00:00.000Z","tags":["instruction-tuned","budget-friendly","long-context"],"benchmarks":[{"source":"openrouter","name":"MMLU","score":82,"date":"2024-07-23T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a70b","name":"Hermes 3 405B Instruct","slug":"hermes-3-405b-instruct","apiIdentifier":"nousresearch/hermes-3-llama-3.1-405b:free","family":"Hermes","version":"3 405B Instruct","description":"Hermes 3 405B Instruct is a large language model from Nous Research. Supports up to 131,072 token context window. Achieves 87.3% on MMLU. Available free with rate limits on OpenRouter.","organizationSlug":"nous-research","organizationName":"Nous Research","contextWindow":131072,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2024-08-16T00:00:00.000Z","tags":["instruction-tuned","long-context"],"benchmarks":[{"source":"openrouter","name":"MMLU","score":87.3,"date":"2024-07-23T00:00:00.000Z"},{"source":"openrouter","name":"HumanEval","score":84.9,"date":"2024-07-23T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6e7","name":"WizardLM-2 8x22B","slug":"wizardlm-2-8x22b","apiIdentifier":"microsoft/wizardlm-2-8x22b","family":"WizardLM-2","version":null,"description":"WizardLM-2 8x22B is a large language model from Microsoft. Supports up to 65,535 token context window. Available from $0.62/M input tokens.","organizationSlug":"microsoft","organizationName":"Microsoft","contextWindow":65535,"maxOutputTokens":8000,"inputPricePerMillion":0.62,"outputPricePerMillion":0.62,"capabilities":["text","code"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2024-04-16T00:00:00.000Z","tags":["budget-friendly"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a3cafdb5cbba71198990659","name":"Segment Anything Model","slug":"segment-anything-model","apiIdentifier":"facebookresearch/segment-anything","family":"Segment Anything Model","version":"1","description":"Segment Anything Model (SAM) is Meta AI Research's promptable image segmentation foundation model for producing object masks from points, boxes, or other prompts in images. The public repository provides inference code, trained checkpoints, and example notebooks for using the model.","organizationSlug":"meta","organizationName":"Meta","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["vision","image-segmentation","promptable-segmentation","research"],"category":"Vision Model","current":true,"deprecated":false,"releaseDate":"2023-04-05T00:00:00.000Z","tags":["image-segmentation","computer-vision","foundation-model","meta-ai","open-source","research-model"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a0bf000fb24974d0528ced4","name":"FinGPT","slug":"fingpt","apiIdentifier":"AI4Finance-Foundation/FinGPT","family":"FinGPT","version":"open-source financial LLM family","description":"FinGPT is an open-source financial large language model project from AI4Finance Foundation. It focuses on finance-specific NLP, sentiment analysis, robo-advisor workflows, technical analysis, and model release workflows through GitHub and Hugging Face.","organizationSlug":"ai4finance-foundation","organizationName":"AI4Finance Foundation","contextWindow":4096,"maxOutputTokens":2048,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","finance-nlp","sentiment-analysis","code","research"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2023-02-11T00:00:00.000Z","tags":["financial-llm","finance","open-source","sentiment-analysis","fingpt"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a5319870add24fdccf21415","name":"CLIP","slug":"clip","apiIdentifier":"openai/clip","family":"CLIP","version":"1","description":"CLIP is OpenAI’s Contrastive Language-Image Pretraining model family, trained to connect images and natural-language text so it can perform zero-shot image classification and image-text retrieval.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["vision","text","zero-shot-classification","image-text-retrieval"],"category":"Vision Model","current":true,"deprecated":false,"releaseDate":"2021-01-05T00:00:00.000Z","tags":["vision-language","openai","zero-shot","image-classification","multimodal","retrieval"],"benchmarks":[{"source":"official-openai-clip-paper","name":"ImageNet zero-shot top-1","score":76.2,"date":"2021-01-05T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6ab5fab5ca0ae1a7e2965837","name":"AgentJev 0.6B","slug":"agent-jev","apiIdentifier":"aimeigaoshou/agent-jev","family":"AgentJev","version":"0.6B","description":"AgentJev 0.6B is a fast decision model for AI agent loops. The GitHub README describes a Qwen3-0.6B-backed model that takes unstructured state plus typed boolean, choice, or score questions and returns probability distributions without decoded output tokens.","organizationSlug":"malevrigns","organizationName":"malevrigns","contextWindow":2048,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["classification","routing","agent-decisioning","probability-scoring"],"category":"Decision Model","current":true,"deprecated":false,"releaseDate":"2026-09-21T00:00:00.000Z","tags":["agent-jev","decision-model","agent-routing","qwen3","open-weights","probability-scoring","ai-agents"],"benchmarks":[{"source":"https://github.com/malevrigns/agent-jev","name":"Typed Decisions top-1 accuracy","score":79.25,"date":"2026-09-25T00:00:00.000Z"},{"source":"https://github.com/malevrigns/agent-jev","name":"Context window tokens","score":2048,"date":"2026-09-25T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6ab5fab4ca0ae1a7e2965836","name":"Mercury 2.5","slug":"mercury-2-5","apiIdentifier":"mercury-2-5","family":"Mercury","version":"2.5","description":"Mercury 2.5 is a proprietary reasoning model from Inception listed by Artificial Analysis. The model page reports a 260,000-token context window, high output speed, and low listed token pricing for a reasoning-capable model.","organizationSlug":"inception","organizationName":"Inception","contextWindow":260000,"maxOutputTokens":null,"inputPricePerMillion":0.04,"outputPricePerMillion":0.15,"capabilities":["text","reasoning","long-context"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-09-08T00:00:00.000Z","tags":["mercury","inception","reasoning-model","long-context","artificial-analysis","proprietary"],"benchmarks":[{"source":"https://artificialanalysis.ai/models/mercury-2-5","name":"Artificial Analysis Intelligence Index","score":12,"date":"2026-09-25T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/mercury-2-5","name":"AA median output speed tokens/sec","score":770.4,"date":"2026-09-25T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/mercury-2-5","name":"AA median time to first chunk seconds","score":3.1,"date":"2026-09-25T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":12.34},{"_id":"6a9e406d66ca1f046dcbba6e","name":"GPT-6 Astra","slug":"gpt-6-astra","apiIdentifier":"gpt-6-astra","family":"GPT","version":"6 Astra","description":"GPT-6 Astra is OpenAI's frontier model for end-to-end work across computer use, browsing, software engineering, cybersecurity, science, long-context tasks, and professional workflows. OpenAI describes it as its most intelligent and aligned model to date.","organizationSlug":"openai","organizationName":"OpenAI","contextWindow":1050000,"maxOutputTokens":128000,"inputPricePerMillion":10,"outputPricePerMillion":50,"capabilities":["text","vision","code","tool-use","computer-use","browsing","long-context","agentic-workflows"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-09-03T00:00:00.000Z","tags":["openai","frontier-model","computer-use","coding","long-context","agentic-ai","api"],"benchmarks":[{"source":"https://openai.com/index/gpt-6-astra/","name":"FrontierMath Tier 4 (v2)","score":97.6,"canonicalName":"FrontierMath Tier 4","rawName":"FrontierMath Tier 4 (v2)","sourceUrl":"https://openai.com/index/gpt-6-astra/","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"GPT-6 Astra","configuration":{"protocol":"OpenAI GPT-6 Astra release evaluation; source footnotes apply"},"benchmarkVersion":"v2","artifactId":"ea74e2fffbc50210","evidenceId":"gpt-6-astra:benchmark:0","confidence":"high","higherIsBetter":true,"date":"2026-09-03T00:00:00.000Z","observedAt":"2026-09-12T16:18:05.575Z"},{"source":"https://openai.com/index/gpt-6-astra/","name":"ARC-AGI-3","score":99.9,"canonicalName":"ARC-AGI-3","rawName":"ARC-AGI-3","sourceUrl":"https://openai.com/index/gpt-6-astra/","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"GPT-6 Astra","configuration":{"protocol":"OpenAI GPT-6 Astra release evaluation; source footnotes apply"},"benchmarkVersion":"3","artifactId":"ea74e2fffbc50210","evidenceId":"gpt-6-astra:benchmark:1","confidence":"high","higherIsBetter":true,"date":"2026-09-03T00:00:00.000Z","observedAt":"2026-09-12T16:18:05.575Z"},{"source":"https://openai.com/index/gpt-6-astra/","name":"Agents' Last Exam","score":59.3,"canonicalName":"Agents' Last Exam","rawName":"Agents' Last Exam","sourceUrl":"https://openai.com/index/gpt-6-astra/","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"GPT-6 Astra","configuration":{"protocol":"OpenAI GPT-6 Astra release evaluation; source footnotes apply"},"artifactId":"ea74e2fffbc50210","evidenceId":"gpt-6-astra:benchmark:2","confidence":"high","higherIsBetter":true,"date":"2026-09-03T00:00:00.000Z","observedAt":"2026-09-12T16:18:05.575Z"},{"source":"https://openai.com/index/gpt-6-astra/","name":"Terminal-Bench 4.0","score":57.9,"canonicalName":"Terminal-Bench","rawName":"Terminal-Bench 4.0","sourceUrl":"https://openai.com/index/gpt-6-astra/","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"GPT-6 Astra","configuration":{"protocol":"OpenAI GPT-6 Astra release evaluation; source footnotes apply"},"benchmarkVersion":"4.0","artifactId":"ea74e2fffbc50210","evidenceId":"gpt-6-astra:benchmark:3","confidence":"high","higherIsBetter":true,"date":"2026-09-03T00:00:00.000Z","observedAt":"2026-09-12T16:18:05.575Z"},{"source":"https://openai.com/index/gpt-6-astra/","name":"GPQA Diamond","score":96,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://openai.com/index/gpt-6-astra/","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"GPT-6 Astra","configuration":{"protocol":"OpenAI GPT-6 Astra release evaluation; source footnotes apply"},"artifactId":"ea74e2fffbc50210","evidenceId":"gpt-6-astra:benchmark:4","confidence":"high","higherIsBetter":true,"date":"2026-09-03T00:00:00.000Z","observedAt":"2026-09-12T16:18:05.575Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":52.8141,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GPT-6 Astra (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"2f339a97-9a0d-499a-9cb5-e0db665bfa25","evaluatorModelSlug":"gpt-6-astra","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"gpt-6-astra:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":96.0606,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-6 Astra (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"2f339a97-9a0d-499a-9cb5-e0db665bfa25","evaluatorModelSlug":"gpt-6-astra","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-6-astra:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":56.4815,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-6 Astra (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"2f339a97-9a0d-499a-9cb5-e0db665bfa25","evaluatorModelSlug":"gpt-6-astra","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"gpt-6-astra:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":88.3895,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GPT-6 Astra (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"2f339a97-9a0d-499a-9cb5-e0db665bfa25","evaluatorModelSlug":"gpt-6-astra","reasoning":true,"effort":"max","selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"gpt-6-astra:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a9ced5e689399496ff97a4e","name":"Manaca 1B Base","slug":"manaca-1b-base","apiIdentifier":"menezesbruno/manaca-1b-base","family":"Manaca","version":"1B Base","description":"Manaca 1B Base is an open, reproducible Brazilian-Portuguese text-generation model built with a Llama-style architecture and published with a Hugging Face model card.","organizationSlug":"instituto-ia-lncc","organizationName":"Instituto IA/LNCC","contextWindow":4096,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","portuguese","text-generation"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":"2026-09-01T00:00:00.000Z","tags":["portuguese","brazilian-portuguese","open-weight","llama","text-generation","transformers"],"benchmarks":[{"source":"https://huggingface.co/menezesbruno/manaca-1b-base","name":"CALAME-PT","score":60.63,"canonicalName":"CALAME-PT","rawName":"CALAME-PT","sourceUrl":"https://huggingface.co/menezesbruno/manaca-1b-base","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Manaca-1B Base (1.72B)","configuration":{"task":"last-word generation","language":"pt-BR"},"artifactId":"75f5b572f97076bd","evidenceId":"manaca-1b-base:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:18:02.316Z"},{"source":"https://huggingface.co/menezesbruno/manaca-1b-base","name":"LAMBADA-PT","score":45.31,"canonicalName":"LAMBADA-PT","rawName":"LAMBADA-PT","sourceUrl":"https://huggingface.co/menezesbruno/manaca-1b-base","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Manaca-1B Base (1.72B)","configuration":{"shots":0,"scoring":"log-likelihood","language":"pt-BR"},"artifactId":"75f5b572f97076bd","evidenceId":"manaca-1b-base:new-benchmark:1","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:18:02.316Z"},{"source":"https://huggingface.co/menezesbruno/manaca-1b-base","name":"HellaSwag-PT","score":41.61,"canonicalName":"HellaSwag-PT","rawName":"HellaSwag-PT","sourceUrl":"https://huggingface.co/menezesbruno/manaca-1b-base","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Manaca-1B Base (1.72B)","configuration":{"shots":10,"scoring":"log-likelihood","language":"pt-BR"},"artifactId":"75f5b572f97076bd","evidenceId":"manaca-1b-base:new-benchmark:2","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:18:02.316Z"},{"source":"https://huggingface.co/menezesbruno/manaca-1b-base","name":"ARC-Challenge-PT","score":27.18,"canonicalName":"ARC-Challenge-PT","rawName":"ARC-Challenge-PT","sourceUrl":"https://huggingface.co/menezesbruno/manaca-1b-base","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"Manaca-1B Base (1.72B)","configuration":{"shots":25,"scoring":"log-likelihood","language":"pt-BR"},"artifactId":"75f5b572f97076bd","evidenceId":"manaca-1b-base:new-benchmark:3","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:18:02.316Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a1d0bcdc03df530171086b7","name":"MOSS-TTS","slug":"moss-tts","apiIdentifier":"https://github.com/OpenMOSS/MOSS-TTS","family":"MOSS-TTS","version":"family","description":"MOSS-TTS Family is an open-source speech and sound generation model family from MOSI.AI and the OpenMOSS team. It covers long-form speech, multi-speaker dialogue, voice and character design, environmental sound effects, and streaming TTS.","organizationSlug":"openmoss","organizationName":"OpenMOSS","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text-to-speech","audio-generation","voice-design","streaming-tts"],"category":"Speech Generation Model","current":true,"deprecated":false,"releaseDate":"2026-02-07T14:05:49.000Z","tags":["open-source","github","ai-model","research","moss-tts"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69f5865c2e3561b8f4a6f687","name":"TimesFM 2.5","slug":"timesfm","apiIdentifier":"google/timesfm-2.5-200m-pytorch","family":"TimesFM","version":"2.5","description":"TimesFM (Time Series Foundation Model) is a pretrained decoder-only foundation model by Google Research for time-series forecasting. It uses a patched-decoder attention architecture where patches of time-points are treated as tokens, enabling zero-shot forecasting across domains and temporal granularities from minutes to yearly.","organizationSlug":"google","organizationName":"Google","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["time-series-forecasting","zero-shot-forecasting","quantile-forecasting"],"category":"Time Series Foundation Model","current":true,"deprecated":false,"releaseDate":"2025-09-15T00:00:00.000Z","tags":["time-series","forecasting","foundation-model","google-research","decoder-only","zero-shot","open-source","pytorch","bigquery","quantile"],"benchmarks":[{"source":"google-research","name":"Monash Archive","score":85,"date":"2025-09-15T00:00:00.000Z"},{"source":"google-research","name":"ETT Long-horizon","score":78,"date":"2025-09-15T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a34c6e708c5191dbb1e0908","name":"RF-DETR","slug":"rf-detr","apiIdentifier":"https://github.com/roboflow/rf-detr","family":"RF-DETR","version":"1.4","description":"RF-DETR is Roboflow’s real-time transformer model family for object detection, instance segmentation, and preview keypoint detection. The source README reports Microsoft COCO and RF100-VL benchmarks and public Apache-licensed model variants.","organizationSlug":"roboflow","organizationName":"Roboflow","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["object-detection","instance-segmentation","keypoint-detection","vision","fine-tuning"],"category":"Vision Model","current":true,"deprecated":false,"releaseDate":"2025-03-19T20:43:00.000Z","tags":["computer-vision","object-detection","segmentation","open-source","roboflow","github","fine-tuning"],"benchmarks":[{"source":"Roboflow RF-DETR README benchmark table","name":"COCO AP50:95 Object Detection","score":60.1,"date":"2026-06-19T00:00:00.000Z"},{"source":"Roboflow RF-DETR README benchmark table","name":"RF100-VL AP50:95 Object Detection","score":63.2,"date":"2026-06-19T00:00:00.000Z"},{"source":"Roboflow RF-DETR README benchmark table","name":"COCO Keypoint AP50:95","score":71.8,"date":"2026-06-19T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a24f4e0e6e20edd09263539","name":"SAM 2","slug":"sam-2","apiIdentifier":"facebook/sam2-hiera-large","family":"Segment Anything Model","version":"2.1","description":"SAM 2, or Segment Anything Model 2, is Meta FAIR’s promptable visual segmentation model for images and videos. The official repository provides inference code, SAM 2.1 checkpoints, example notebooks, training resources, and benchmark tables for SA-V, MOSE, and LVOS video object segmentation tasks.","organizationSlug":"meta","organizationName":"Meta","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["vision","image-segmentation","video-segmentation","promptable-segmentation","open-weights"],"category":"Image Model","current":true,"deprecated":false,"releaseDate":"2024-07-29T00:00:00.000Z","tags":["sam-2","segment-anything","image-segmentation","video-segmentation","meta","fair","open-weights","computer-vision"],"benchmarks":[{"source":"official SAM 2 README, sam2.1_hiera_large","name":"SA-V test J&F","score":79.5,"canonicalName":"SA-V test J&F","rawName":"SA-V test J&F","sourceUrl":"https://github.com/facebookresearch/sam2","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"sam2.1_hiera_large","configuration":{"checkpoint":"SAM 2.1 Hiera Large","hardware":"A100","sourceTable":"Official README SAM 2.1 table"},"artifactId":"e55fa4bef1d95bb2","evidenceId":"sam-2:benchmark:0","confidence":"high","higherIsBetter":true,"date":"2024-09-29T00:00:00.000Z","observedAt":"2026-09-12T16:17:46.028Z"},{"source":"official SAM 2 README, sam2.1_hiera_large","name":"MOSE val J&F","score":74.6,"canonicalName":"MOSE val J&F","rawName":"MOSE val J&F","sourceUrl":"https://github.com/facebookresearch/sam2","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"sam2.1_hiera_large","configuration":{"checkpoint":"SAM 2.1 Hiera Large","hardware":"A100","sourceTable":"Official README SAM 2.1 table"},"artifactId":"e55fa4bef1d95bb2","evidenceId":"sam-2:benchmark:1","confidence":"high","higherIsBetter":true,"date":"2024-09-29T00:00:00.000Z","observedAt":"2026-09-12T16:17:46.028Z"},{"source":"official SAM 2 README, sam2.1_hiera_large","name":"LVOS v2 J&F","score":80.6,"canonicalName":"LVOS v2 J&F","rawName":"LVOS v2 J&F","sourceUrl":"https://github.com/facebookresearch/sam2","sourceType":"first-party-model-card","unit":"percent","scale":"0-100","modelVariant":"sam2.1_hiera_large","configuration":{"checkpoint":"SAM 2.1 Hiera Large","hardware":"A100","sourceTable":"Official README SAM 2.1 table"},"artifactId":"e55fa4bef1d95bb2","evidenceId":"sam-2:benchmark:2","confidence":"high","higherIsBetter":true,"date":"2024-09-29T00:00:00.000Z","observedAt":"2026-09-12T16:17:46.028Z"},{"source":"official SAM 2 README, sam2.1_hiera_large","name":"Speed FPS A100","score":39.5,"canonicalName":"Speed FPS A100","rawName":"Speed FPS A100","sourceUrl":"https://github.com/facebookresearch/sam2","sourceType":"first-party-model-card","unit":"frames/s","scale":"nonnegative","modelVariant":"sam2.1_hiera_large","configuration":{"checkpoint":"SAM 2.1 Hiera Large","hardware":"A100","sourceTable":"Official README SAM 2.1 table"},"artifactId":"e55fa4bef1d95bb2","evidenceId":"sam-2:benchmark:3","confidence":"high","higherIsBetter":true,"date":"2024-09-29T00:00:00.000Z","observedAt":"2026-09-12T16:17:46.028Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a1d0bcfc03df530171086b8","name":"Eagle","slug":"eagle","apiIdentifier":"https://github.com/NVlabs/Eagle","family":"Eagle","version":"research-release","description":"Eagle is an NVLabs frontier vision-language model project focused on data-centric strategies for multimodal image and language understanding.","organizationSlug":"nvidia","organizationName":"NVIDIA","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["vision","image-understanding","multimodal","research"],"category":"Vision-Language Model","current":true,"deprecated":false,"releaseDate":"2024-06-27T07:22:52.000Z","tags":["open-source","github","ai-model","research","eagle"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6ab89ced2cdbeb4cb7c5fbd8","name":"Bark","slug":"bark","apiIdentifier":"suno-ai/bark","family":"Bark","version":"v0","description":"Bark is Suno’s open-source transformer-based text-to-audio model. It generates realistic multilingual speech plus music-like audio, background noise, sound effects, and nonverbal vocal cues such as laughing or sighing.","organizationSlug":"suno-ai","organizationName":"Suno","contextWindow":256,"maxOutputTokens":null,"inputPricePerMillion":0,"outputPricePerMillion":0,"capabilities":["text-to-audio","speech-generation","multilingual-speech","sound-effects","voice-cloning-style-prompts"],"category":"Audio Model","current":true,"deprecated":false,"releaseDate":"2023-04-20T00:00:00.000Z","tags":["text-to-audio","speech-synthesis","audio-generation","open-source","suno"],"benchmarks":[{"source":"Suno Bark README","name":"Open-source checkpoint availability","score":1,"date":"2023-05-01T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a9e406a66ca1f046dcbba6c","name":"Stable Diffusion","slug":"stable-diffusion","apiIdentifier":"CompVis/stable-diffusion-v1","family":"Stable Diffusion","version":"v1","description":"Stable Diffusion is a latent text-to-image diffusion model from CompVis, developed with compute support from Stability AI and support from LAION. The v1 model generates and modifies images from text prompts using a latent diffusion architecture and a frozen CLIP ViT-L/14 text encoder.","organizationSlug":"compvis","organizationName":"CompVis","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text-to-image","image-generation","image-to-image","research"],"category":"Image Model","current":true,"deprecated":false,"releaseDate":"2022-08-22T00:00:00.000Z","tags":["text-to-image","diffusion","latent-diffusion","image-generation","open-weights","research","creative-ai"],"benchmarks":[{"source":"https://openaccess.thecvf.com/content/CVPR2022/html/Rombach_High-Resolution_Image_Synthesis_With_Latent_Diffusion_Models_CVPR_2022_paper.html","name":"CVPR 2022 latent diffusion paper","score":null,"date":"2022-06-01T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a34c6d9ce54e151f04ae76c","name":"GLM-5.2","slug":"glm-5-2","apiIdentifier":"glm-5.2-max","family":"GLM","version":"5.2 max","description":"GLM-5.2 (max) is a Z AI text model tracked by Artificial Analysis. The page reports a 1 million token context window, text input and text output, a 51.09 Artificial Analysis Intelligence Index, and measured output speed around 102 tokens per second.","organizationSlug":"z-ai","organizationName":"Z.ai","contextWindow":1000000,"maxOutputTokens":1000000,"inputPricePerMillion":1.4,"outputPricePerMillion":4.4,"capabilities":["text","reasoning","long-context"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":null,"tags":["glm","z-ai","long-context","reasoning","text-model","artificial-analysis","llm"],"benchmarks":[{"source":"Artificial Analysis model page","name":"Artificial Analysis Intelligence Index v4.1","score":51.0858347714416,"date":"2026-06-19T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/glm-5-2","name":"Artificial Analysis Intelligence Index v4.3","score":34,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/glm-5-2","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GLM-5.2 (max)","configuration":{"protocol":"Artificial Analysis Intelligence Index v4.3; exact effort/fallback in model variant"},"benchmarkVersion":"4.3","artifactId":"93b45db1a8f8895c","evidenceId":"glm-5-2:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:17:46.700Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":34.0052,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"GLM-5.2 (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"f7a4ea75-e548-4069-80d4-9be8bc7c009b","evaluatorModelSlug":"glm-5-2","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"glm-5-2:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":89.4949,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5.2 (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"f7a4ea75-e548-4069-80d4-9be8bc7c009b","evaluatorModelSlug":"glm-5-2","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"glm-5-2:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":51.1574,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5.2 (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"f7a4ea75-e548-4069-80d4-9be8bc7c009b","evaluatorModelSlug":"glm-5-2","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"glm-5-2:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":77.9026,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"GLM-5.2 (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"f7a4ea75-e548-4069-80d4-9be8bc7c009b","evaluatorModelSlug":"glm-5-2","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"glm-5-2:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":51.0858347714416},{"_id":"6a07f3621c42579a110320f4","name":"MiniCPM","slug":"minicpm","apiIdentifier":"OpenBMB/MiniCPM","family":"MiniCPM","version":"4 / 4.1","description":"MiniCPM is OpenBMB’s ultra-efficient open language-model family for edge and end-device deployment. The MiniCPM4 and MiniCPM4.1 lines focus on fast local reasoning, while MiniCPM-SALA extends the family toward sparse/linear attention and million-token context research.","organizationSlug":"openbmb","organizationName":"OpenBMB","contextWindow":128000,"maxOutputTokens":32768,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text","code","reasoning","local-inference"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":null,"tags":["open-weight","edge-ai","local-llm","reasoning","openbmb"],"benchmarks":[{"source":"official-github-readme","name":"MiniCPM-SALA standard benchmark average","score":76.53,"date":"2026-05-16T04:34:12.133Z"},{"source":"official-github-readme","name":"MiniCPM-SALA long-context average","score":38.97,"date":"2026-05-16T04:34:12.133Z"},{"source":"official-github-readme","name":"MiniCPM-SALA 2048K extrapolation score","score":81.6,"date":"2026-05-16T04:34:12.133Z"},{"source":"official-github-readme","name":"MiniCPM4.1 reasoning decoding speedup","score":3,"date":"2026-05-16T04:34:12.133Z"},{"source":"official-github-readme","name":"MiniCPM4 Jetson AGX Orin decoding speedup vs Qwen3-8B","score":7,"date":"2026-05-16T04:34:12.133Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a6de5d66d8067980a992c77","name":"Kimi K3","slug":"kimi-k3","apiIdentifier":"kimi-k3","family":"Kimi","version":"K3","description":"Kimi K3 is Moonshot AI open-weight 2.8T-parameter flagship model for long-horizon coding, end-to-end knowledge work, reasoning, and native multimodal understanding. Official Kimi documentation lists the API model ID kimi-k3, a 1M-token context window, automatic context caching, tool calls, JSON mode, structured outputs, and always-on reasoning effort controls.","organizationSlug":"moonshot-ai","organizationName":"Moonshot AI","contextWindow":1048576,"maxOutputTokens":131072,"inputPricePerMillion":3,"outputPricePerMillion":15,"capabilities":["text","vision","reasoning","coding","tool-use","structured-output","json-mode","long-context","agentic"],"category":"Large Language Model","current":true,"deprecated":false,"releaseDate":null,"tags":["reasoning","coding","long-context","multimodal","open-weight","agentic"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index","score":57,"rawName":"Artificial Analysis Intelligence Index","sourceUrl":"https://artificialanalysis.ai/articles/kimi-k3-agentic-knowledge-benchmark","sourceType":"benchmark-leaderboard","unit":"index","scale":"0-100","evidenceId":"aa-kimi-k3-briefcase","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-08-01T00:00:00.000Z"},{"source":"artificial-analysis","name":"AA-Briefcase","score":1543,"rawName":"AA-Briefcase Elo","sourceUrl":"https://artificialanalysis.ai/articles/kimi-k3-agentic-knowledge-benchmark","sourceType":"benchmark-leaderboard","unit":"Elo","scale":"Elo","evidenceId":"aa-kimi-k3-briefcase","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-08-01T00:00:00.000Z"},{"source":"artificial-analysis","name":"AA-Briefcase Rubric Pass Rate","score":51,"rawName":"AA-Briefcase rubric pass rate","sourceUrl":"https://artificialanalysis.ai/articles/kimi-k3-agentic-knowledge-benchmark","sourceType":"benchmark-leaderboard","unit":"percent","scale":"0-100","evidenceId":"aa-kimi-k3-briefcase","confidence":"high","date":"2026-07-21T00:00:00.000Z","observedDate":"2026-08-01T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1.1","score":60,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1.1","sourceUrl":"https://artificialanalysis.ai/models/kimi-k3","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"Kimi K3 (max)","evidenceId":"sb-aa-kimi-k3-model-20260808","confidence":"high","date":"2026-08-06T00:00:00.000Z","observedDate":"2026-08-08T00:00:00.000Z"},{"source":"https://artificialanalysis.ai/models/kimi-k3","name":"Artificial Analysis Intelligence Index v4.3","score":44,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/kimi-k3","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Kimi K3 (max)","configuration":{"protocol":"Artificial Analysis Intelligence Index v4.3; exact effort/fallback in model variant"},"benchmarkVersion":"4.3","artifactId":"5d65f998595ae7d5","evidenceId":"kimi-k3:new-benchmark:0","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T16:28:36.055Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":43.7842,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"Kimi K3 (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"f7d2fc3e-1f7b-405f-818c-07952a4af78f","evaluatorModelSlug":"kimi-k3","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"kimi-k3:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":93.5354,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Kimi K3 (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"f7d2fc3e-1f7b-405f-818c-07952a4af78f","evaluatorModelSlug":"kimi-k3","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"kimi-k3:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":59.4907,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Kimi K3 (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"f7d2fc3e-1f7b-405f-818c-07952a4af78f","evaluatorModelSlug":"kimi-k3","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"kimi-k3:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":85.0187,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Kimi K3 (max)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"f7d2fc3e-1f7b-405f-818c-07952a4af78f","evaluatorModelSlug":"kimi-k3","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"kimi-k3:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":60},{"_id":"6aa5b427d47248953e07ce9d","name":"FLUX.2 [klein] 4B","slug":"flux-2-klein-4b","apiIdentifier":null,"family":"FLUX.2","version":"klein 4B distilled","description":"FLUX.2 klein 4B is a distilled image-generation and multi-reference editing model developed by Black Forest Labs. The official checkpoint has 4 billion parameters and is released under Apache 2.0. Its model card reports approximately 13GB VRAM for local use; requirements depend on the runtime and configuration. It supports text-to-image and image-to-image workflows. Generated text may be distorted, and outputs can miss the prompt or reflect biases. This record refers to black-forest-labs/FLUX.2-klein-4B, not the base checkpoint or klein 9B. Source: the official Black Forest Labs Hugging Face model card, checked September 12, 2026.","organizationSlug":"black-forest-labs","organizationName":"Black Forest Labs","contextWindow":null,"maxOutputTokens":null,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["text-to-image","image-to-image","multi-reference editing"],"category":"Image Model","current":true,"deprecated":false,"releaseDate":null,"tags":["image generation","image editing","multi-reference","open weights","Apache 2.0","distilled"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"6a732c0f3bc1ce213a0a3339","name":"DeepSeek V4 Flash 0731","slug":"deepseek-v4-flash-0731","apiIdentifier":"deepseek-v4-flash","family":"DeepSeek V4 Flash","version":"0731","description":"DeepSeek V4 Flash 0731 is an open-weight, cost-efficient reasoning model for long-context coding, agent tasks, and API workflows through DeepSeek.","organizationSlug":"deepseek","organizationName":"DeepSeek","contextWindow":1048576,"maxOutputTokens":393216,"inputPricePerMillion":null,"outputPricePerMillion":null,"capabilities":["reasoning","coding","tool-use","structured-output","long-context","responses-api"],"category":"Large Language Model","current":false,"deprecated":true,"releaseDate":"2026-07-31T00:00:00.000Z","tags":["reasoning","open-weights","coding","long-context","responses-api"],"benchmarks":[{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1","score":50,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1","sourceUrl":"https://artificialanalysis.ai/models/deepseek-v4-flash","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"DeepSeek V4 Flash 0731 (Reasoning, Max Effort)","evidenceId":"sb-aa-deepseek-v4-flash-model","confidence":"high","date":"2026-08-05T00:00:00.000Z","observedDate":"2026-08-05T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.1.1","score":52,"canonicalName":"Artificial Analysis Intelligence Index","rawName":"Artificial Analysis Intelligence Index v4.1.1","sourceUrl":"https://artificialanalysis.ai/models/deepseek-v4-flash","sourceType":"independent-benchmark","unit":"index","scale":"0-100 higher is better","modelVariant":"DeepSeek V4 Flash 0731 (Reasoning, Max Effort)","evidenceId":"sb-aa-deepseek-v4-flash-20260807","confidence":"high","date":"2026-08-06T00:00:00.000Z","observedDate":"2026-08-07T00:00:00.000Z"},{"source":"artificial-analysis","name":"Artificial Analysis Intelligence Index v4.3","score":34.5282,"canonicalName":"Artificial Analysis Intelligence Index v4.3","rawName":"Artificial Analysis Intelligence Index v4.3","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"index","scale":"AA Intelligence Index v4.3","modelVariant":"DeepSeek V4 Flash 0731 (Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"fe4c0848-e284-4e52-a79d-cdc28392f1a9","evaluatorModelSlug":"deepseek-v4-flash","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"4.3","artifactId":"aa-common-cohort","evidenceId":"deepseek-v4-flash-0731:aa:intelligenceIndex","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"GPQA Diamond","score":90.8081,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V4 Flash 0731 (Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"fe4c0848-e284-4e52-a79d-cdc28392f1a9","evaluatorModelSlug":"deepseek-v4-flash","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"deepseek-v4-flash-0731:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"SciCode","score":50.3472,"canonicalName":"SciCode","rawName":"SciCode","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V4 Flash 0731 (Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","evaluation":"AA SciCode implementation","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"fe4c0848-e284-4e52-a79d-cdc28392f1a9","evaluatorModelSlug":"deepseek-v4-flash","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"deepseek-v4-flash-0731:aa:scicode","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":78.6517,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"DeepSeek V4 Flash 0731 (Reasoning, Max Effort)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"fe4c0848-e284-4e52-a79d-cdc28392f1a9","evaluatorModelSlug":"deepseek-v4-flash","reasoning":true,"effort":"max","selection":"existing exact source mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"deepseek-v4-flash-0731:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":52},{"_id":"69e45e9faccbb529d120a6c3","name":"MiMo-V2-Omni","slug":"mimo-v2-omni","apiIdentifier":"xiaomi/mimo-v2-omni","family":"MiMo","version":"-V2-Omni","description":"MiMo-V2-Omni is a multimodal llm from Xiaomi. Supports up to 262,144 token context window. Available from $0.40/M input tokens.","organizationSlug":"xiaomi","organizationName":"Xiaomi","contextWindow":262144,"maxOutputTokens":65536,"inputPricePerMillion":0.4,"outputPricePerMillion":2,"capabilities":["text","vision","audio","video","code"],"category":"Multimodal LLM","current":false,"deprecated":true,"releaseDate":"2026-03-18T00:00:00.000Z","tags":["multimodal","audio","video","budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":82.8283,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"MiMo-V2-Omni","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"94d09368-9035-47cf-963a-b4310b433a16","evaluatorModelSlug":"mimo-v2-omni","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"mimo-v2-omni:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1387,"arenaVotes":10994,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6e3","name":"MiMo-V2-Pro","slug":"mimo-v2-pro","apiIdentifier":"xiaomi/mimo-v2-pro","family":"MiMo","version":"-V2-Pro","description":"MiMo-V2-Pro is a large language model from Xiaomi. Supports up to 1,048,576 token context window. Available from $1.00/M input tokens.","organizationSlug":"xiaomi","organizationName":"Xiaomi","contextWindow":1048576,"maxOutputTokens":131072,"inputPricePerMillion":1,"outputPricePerMillion":3,"capabilities":["text","code"],"category":"Large Language Model","current":false,"deprecated":true,"releaseDate":"2026-03-18T00:00:00.000Z","tags":["long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":86.9697,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"MiMo-V2-Pro","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"5d8183dc-24f4-46c5-a1d0-d937de149364","evaluatorModelSlug":"mimo-v2-pro","reasoning":true,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"mimo-v2-pro:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1446,"arenaVotes":10116,"aaIntelligenceIndex":49},{"_id":"69e45e9faccbb529d120a6d2","name":"Mistral Small Creative","slug":"mistral-small-creative","apiIdentifier":"mistralai/mistral-small-creative","family":"Mistral","version":"Creative","description":"Mistral Small Creative is a large language model from Mistral AI. Supports up to 32,768 token context window. Achieves 76.5% on MMLU. Available from $0.10/M input tokens.","organizationSlug":"mistral","organizationName":"Mistral AI","contextWindow":32768,"maxOutputTokens":null,"inputPricePerMillion":0.1,"outputPricePerMillion":0.3,"capabilities":["text","code","tool-use"],"category":"Large Language Model","current":false,"deprecated":true,"releaseDate":"2025-12-16T00:00:00.000Z","tags":["budget-friendly"],"benchmarks":[{"source":"mistral","name":"MMLU","score":81,"date":"2025-03-05T00:00:00.000Z"}],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6fd","name":"MiMo-V2-Flash","slug":"mimo-v2-flash","apiIdentifier":"xiaomi/mimo-v2-flash","family":"MiMo","version":"-V2-Flash","description":"MiMo-V2-Flash is a large language model from Xiaomi. Supports up to 262,144 token context window. Available from $0.09/M input tokens.","organizationSlug":"xiaomi","organizationName":"Xiaomi","contextWindow":262144,"maxOutputTokens":65536,"inputPricePerMillion":0.09,"outputPricePerMillion":0.29,"capabilities":["text","code"],"category":"Large Language Model","current":false,"deprecated":true,"releaseDate":"2025-12-14T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":65.5556,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"MiMo-V2-Flash (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"82b36b4d-84dd-4bc0-ad32-e3aee9442789","evaluatorModelSlug":"mimo-v2-flash","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"mimo-v2-flash:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"Terminal-Bench 2.1","score":61.7978,"canonicalName":"Terminal-Bench 2.1","rawName":"Terminal-Bench 2.1","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"MiMo-V2-Flash (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","harness":"Terminus 2","tasks":89,"repeats":3,"metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"82b36b4d-84dd-4bc0-ad32-e3aee9442789","evaluatorModelSlug":"mimo-v2-flash","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2.1","artifactId":"aa-common-cohort","evidenceId":"mimo-v2-flash:aa:terminalbenchV21","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":67.6667,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"MiMo-V2-Flash (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"82b36b4d-84dd-4bc0-ad32-e3aee9442789","evaluatorModelSlug":"mimo-v2-flash","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"mimo-v2-flash:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1392,"arenaVotes":31981,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6ec","name":"Devstral 2 2512","slug":"devstral-2-2512","apiIdentifier":"mistralai/devstral-2512","family":"Devstral","version":null,"description":"Devstral 2 2512 is a large language model from Mistral AI. Supports up to 262,144 token context window. Available from $0.40/M input tokens.","organizationSlug":"mistral","organizationName":"Mistral AI","contextWindow":262144,"maxOutputTokens":null,"inputPricePerMillion":0.4,"outputPricePerMillion":2,"capabilities":["text","code","tool-use"],"category":"Large Language Model","current":false,"deprecated":true,"releaseDate":"2025-12-09T00:00:00.000Z","tags":["budget-friendly","long-context"],"benchmarks":[],"arenaScore":null,"arenaVotes":null,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a6f1","name":"Grok 4.1 Fast","slug":"grok-41-fast","apiIdentifier":"x-ai/grok-4.1-fast","family":"Grok","version":"Fast","description":"Grok 4.1 Fast is a multimodal llm from xAI. Supports up to 2,000,000 token context window. Available from $0.20/M input tokens.","organizationSlug":"xai","organizationName":"xAI","contextWindow":2000000,"maxOutputTokens":30000,"inputPricePerMillion":0.2,"outputPricePerMillion":0.5,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":false,"deprecated":true,"releaseDate":"2025-11-19T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":63.7374,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Grok 4.1 Fast (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"49fd01f9-887d-4479-b8ce-771a81ecef4e","evaluatorModelSlug":"grok-4-1-fast","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"grok-41-fast:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":34.3333,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Grok 4.1 Fast (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"49fd01f9-887d-4479-b8ce-771a81ecef4e","evaluatorModelSlug":"grok-4-1-fast","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"grok-41-fast:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1432,"arenaVotes":44414,"aaIntelligenceIndex":null},{"_id":"69e45e9faccbb529d120a707","name":"Grok 4 Fast","slug":"grok-4-fast","apiIdentifier":"x-ai/grok-4-fast","family":"Grok","version":"Fast","description":"Grok 4 Fast is a multimodal llm from xAI. Supports up to 2,000,000 token context window. Available from $0.20/M input tokens.","organizationSlug":"xai","organizationName":"xAI","contextWindow":2000000,"maxOutputTokens":30000,"inputPricePerMillion":0.2,"outputPricePerMillion":0.5,"capabilities":["text","vision","code"],"category":"Multimodal LLM","current":false,"deprecated":true,"releaseDate":"2025-09-19T00:00:00.000Z","tags":["multimodal","budget-friendly","long-context"],"benchmarks":[{"source":"artificial-analysis","name":"GPQA Diamond","score":60.6061,"canonicalName":"GPQA Diamond","rawName":"GPQA Diamond","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Grok 4 Fast (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"Diamond (198 questions)","metric":"pass@1","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"2dbb6dc7-8c40-4b6d-af9c-cf805f83b79a","evaluatorModelSlug":"grok-4-fast","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"artifactId":"aa-common-cohort","evidenceId":"grok-4-fast:aa:gpqa","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"},{"source":"artificial-analysis","name":"AIME 2025","score":41.3333,"canonicalName":"AIME 2025","rawName":"AIME 2025","sourceUrl":"https://artificialanalysis.ai/models/claude-opus-5","sourceType":"independent-benchmark","unit":"percent","scale":"0-100","modelVariant":"Grok 4 Fast (Non-reasoning)","configuration":{"protocol":"aa-independent-2026-09-v4.3","subset":"AIME I and II 2025","repeats":10,"metric":"pass@1","status":"retired from active AA reporting","methodologyUrl":"https://artificialanalysis.ai/methodology/intelligence-benchmarking","modelSettings":{"evaluatorModelId":"2dbb6dc7-8c40-4b6d-af9c-cf805f83b79a","evaluatorModelSlug":"grok-4-fast","reasoning":false,"effort":null,"selection":"unique canonical release mapping","caveat":"Named evaluated configuration; not an equal-compute or equal-reasoning-budget comparison"}},"benchmarkVersion":"2025","artifactId":"aa-common-cohort","evidenceId":"grok-4-fast:aa:aime25","confidence":"high","higherIsBetter":true,"date":null,"observedAt":"2026-09-12T17:34:07.079Z"}],"arenaScore":1404,"arenaVotes":18750,"aaIntelligenceIndex":null}]}