LLM Comparison
DeepSeek V3.2 vs DeepSeek V3.2 Exp
Side-by-side specs, pricing & capabilities
Add to comparison
2/6 models| Organization | ||
| OpenTools Score | ||
| Family | DeepSeek | DeepSeek |
| Status | Current | Current |
| Release Date | Dec 2025 | Sep 2025 |
| Context Window | 164K tokens | 164K tokens |
| Input Price | $0.26/M tokens | $0.27/M tokens |
| Output Price | $0.42/M tokens | $0.41/M tokens |
| Pricing Notes | Cache read: $0.1350/M tokens | — |
| Capabilities | textcode | textcode |
| Max Output | 164K tokens | 66K tokens |
| API Identifier | deepseek/deepseek-v3.2 | deepseek/deepseek-v3.2-exp |
| Benchmarks | ||
| AIME 2025 | 59%artificial-analysis | 57.7%artificial-analysis |
| AIME 2025 | — | 89.3%deepseek |
| BrowseComp | — | 40.1%deepseek |
| GPQA Diamond | 75.1%artificial-analysis | 73.8%artificial-analysis |
| GPQA Diamond | — | 79.9%deepseek |
| HMMT 2025 | — | 83.6%deepseek |
| MMLU-Pro | — | 85%deepseek |
| SWE-bench Verified | — | 67.8%deepseek |
| View DeepSeek V3.2 | View DeepSeek V3.2 Exp | |
Cost Calculator
Enter your expected monthly token usage to compare costs.
| Model | Input | Output | Total / mo | vs Best |
|---|---|---|---|---|
| DeepSeek V3.2Cheapest | $0.26 | $0.21 | $0.47 | — |
| DeepSeek V3.2 Exp | $0.27 | $0.21 | $0.48 | +1% |
Compare hosting configurations
Compare two exact setups, including different configurations of the same model. Hardware fit, measured speed and costs are separate decisions.
Add your cost assumptions (optional)
Use hours for the same scenario period. Cloud billing includes idle time; active hours alone do not set the bill. Local electricity uses whole-system attributable power, powered hours and your tariff. Missing inputs remain unknown.
Measured speed needs the same pinned workload protocol, tokenizer, context, batching and cache state.
| Configuration | Convert and host complete DeepSeek V3.2 DSA reasoning on a 512 GiB CPU workstation | Convert and host complete DeepSeek V3.2 Exp DSA on a 512 GiB CPU workstation |
|---|---|---|
| Model | DeepSeek V3.2 | DeepSeek V3.2 Exp |
| Hardware | Intel Xeon w5-2465X / ASUS W790E-SAGE SE, 512 GiB RAM | Intel Xeon w5-2465X / ASUS W790E-SAGE SE, 512 GiB RAM |
| Runtime | llama.cpp native V3.2 CPU DSA with full reader conversion v0.7.0 (620d2225; full pin in source URL) · CPU | llama.cpp complete DEEPSEEK32 CPU DSA with native Exp conversion v0.7.0 (620d2225; full pin in source URL) · CPU |
| Artifact | Complete native DeepSeek V3.2, reader-generated Q4_K_M with DSA and NextN · Reader-generated full Q4_K_M from original ordinary V3.2 mixed FP8/BF16/F32 through F16; complete experts and DSA/MLA retained · Complete reader-generated Q4_K_M through F16 from original FP8 checkpoint; all163 native shards, main DSA/MLA/experts and NextN retained | Complete DeepSeek V3.2 Exp, reader-generated Q4_K_M with DSA and NextN · Reader-generated full Q4_K_M from original mixed FP8/BF16/F32 native Exp checkpoint through a F16 intermediate; no active-only subset · Reader-native FP8 dequantization to F16 intermediate, then mixed Q4_K_M; all163 original checkpoint shards and main/NextN tensors retained |
| Context / concurrency | 8,192 tokens / 1 | 8,192 tokens / 1 |
| Required memory | Unknown (unknown) | Unknown (unknown) |
| Evidence | Documented setup | Documented setup |
| Hardware fit | Hardware fit unknown No measured full CPU process peak for all quantized weights, DSA/MLA cache, scratch, loading, runtime and OS. Enter available memory per device or unified memory pool. Host RAM: 512 GiB is a planning allocation, not an observed minimum or fit. Free disk space: 4096 GiB (4 TiB) is the prescribed working-volume allocation. The initial download guard requires at least 3.5 TiB free; neither amount is a measured conversion or serving peak. | Hardware fit unknown No measured full CPU process peak for all quantized weights, DSA/MLA cache, scratch, loading, runtime and OS. Enter available memory per device or unified memory pool. Host RAM: 512 GiB is a planning allocation, not an observed minimum or fit. Free disk space: 4096 GiB free working storage is an author prescription, not a measured conversion or serving peak. |
| One-time cost | Unknown hardware: Existing desktop purchase price and incremental hardware cost depend on the reader; unknown. setup: Initial provisioning/build time and labor cost are unknown. | Unknown hardware: Existing desktop purchase price and incremental hardware cost depend on the reader; unknown. setup: Initial provisioning/build time and labor cost are unknown. |
| Recurring scenario cost | Unknown electricity: Actual wall power, electricity rate and powered hours are unknown; no TDP-to-wall-power conversion. storage: Incremental local disk cost unknown;4096 GiB is prescribed working capacity. egress: Data transfer price, allowances and traffic volume are unknown. administration: Maintenance labor rate and hours are unknown. One-time setup and recurring costs are separate. Unknown components are not treated as free. No monthly token output is inferred from peak throughput. | Unknown electricity: Actual wall power, electricity rate and powered hours are unknown; no TDP-to-wall-power conversion. storage: Incremental local disk cost unknown;4096 GiB is prescribed working capacity. egress: Data transfer price, allowances and traffic volume are unknown. administration: Maintenance labor rate and hours are unknown. One-time setup and recurring costs are separate. Unknown components are not treated as free. No monthly token output is inferred from peak throughput. |
| Published measurements (unranked) | No compatible measurement available | No compatible measurement available |
Different currencies and incomplete costs are not ranked. Model quality benchmarks remain on the model pages; they do not establish quality or speed for every deployment.
DeepSeek
DeepSeek V3.2
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.
DeepSeek
DeepSeek V3.2 Exp
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.
More Comparisons
Looking for more AI models?
Browse All LLMs