← Models
Kimi K3
Moonshot AI · open weight · kimi-k3
Kimi K3 is Moonshot AI's flagship open Mixture-of-Experts model for long-horizon coding, knowledge work, and reasoning. It has 2.8 trillion total parameters, activates 16 of 896 experts, and combines Kimi Delta Attention, Attention Residuals, and Stable LatentMoE. The model supports a 1M-token context window, native image and video understanding, tool calling, structured output, and always-on reasoning with max thinking effort at launch.
Benchmark scores
| Benchmark | Score |
|---|---|
| AA-Briefcase | 1548.00 |
| MathVision | 0.98 |
| DeepSearchQA | 0.95 |
| GPQA | 0.94 |
| CharXiv-R | 0.91 |
| BrowseComp | 0.91 |
| OmniDocBench | 0.91 |
| Terminal-Bench 2.1 | 0.88 |
| BabyVision | 0.86 |
| MCP Atlas | 0.84 |
| MMMU-Pro | 0.82 |
| FrontierSWE | 0.81 |
| Program Bench | 0.78 |
| DECK-Bench | 0.73 |
| Toolathlon | 0.73 |
| Kimi Code Bench v2 | 0.73 |
| Humanity's Last Exam | 0.56 |
Pricing
- DeepInfra$2.85 / $14.25
- Novita$3.00 / $15.00
- Fireworks$3.00 / $15.00
- Moonshot AI$3.00 / $15.00
- Together$3.00 / $15.00
Input / output per 1M tokens
AA metrics
No Artificial Analysis link yet.
Arena Elo
- No Arena snapshot linked.