81
OpenTools Scorenormalized /100
9.0
Valuescore / $M
Context1.0M tokens
Price (In / Out)$3.00/M / $15.00/M
CategoryLarge Language Model
Max Output131.1K tokens
About Kimi K3
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.
Capabilities
textvisionreasoningcodingtool usestructured outputjson modelong contextagentic
Input Modalities
textimage
Output Modalities
text
Technical Details
- API Identifier
- kimi-k3
- Category
- Large Language Model
- Context Window
- 1,048,576 tokens
- Max Output Tokens
- 131,072 tokens
Tags
reasoningcodinglong-contextmultimodalopen-weightagentic
Run this model yourself
Local and cloud setup for Kimi K3, with exact configurations, sources and clear evidence limits.
No reviewed setup guide is available for this model yet. Explore configurations with documented support in the hardware finder.
Benchmarks
Performance scores for Kimi K3 across standard benchmarks.
Artificial Analysis Intelligence IndexArtificial Analysis · Jul 2026
57 index
AA-BriefcaseArtificial Analysis · Jul 2026
1543 elo
AA-Briefcase Rubric Pass RateArtificial Analysis · Jul 2026
51%
Artificial Analysis Intelligence Index v4.1.1Artificial Analysis · Aug 2026
60 index
Artificial Analysis Intelligence Index v4.3https://artificialanalysis.ai/models/kimi-k3
44 index
Artificial Analysis Intelligence Index v4.3Artificial Analysis
43.8 index
GPQA DiamondArtificial Analysis
93.5%
SciCodeArtificial Analysis
59.5%
Terminal-Bench 2.1Artificial Analysis
85%
Pricing
Token pricing for Kimi K3 API usage.
Input Tokens
$3.00/M
per million tokens
Output Tokens
$15.00/M
per million tokens
Official Kimi pricing lists kimi-k3 at $3.00 per 1M cache-miss input tokens, $0.30 per 1M cache-hit input tokens, and $15.00 per 1M output tokens. Official troubleshooting docs say max_completion_tokens defaults to 131,072 and maximum output length is 1024*1024 minus prompt tokens.
Moonshot AIai lab
Long-context Kimi models and AI assistant platform
5 ModelsFounded 202313th Floor, Building 1, JD Technology Building, No. 76 Zhichun Road, Haidian District, Beijing, China
View full profileAI Models by Moonshot AI
AI models from the same organization.
| Model | Context Window | Price (In / Out per M) |
|---|---|---|
| Kimi K2.5Current | 262K | $0.38 / $1.72 |
| Kimi K2 ThinkingCurrent | 262K | $0.60 / $2.50 |
| Kimi K2 0905Current | 262K | $0.40 / $2.00 |
| Kimi K2 0711Current | 131K | $0.57 / $2.30 |