Readme
1. Model Introduction
Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated parameters and 1 trillion total parameters. Trained with the Muon optimizer, Kimi K2 achieves exceptional performance across frontier knowledge, reasoning, and coding tasks while being meticulously optimized for agentic capabilities.
Key Features
- Large-Scale Training: Pre-trained a 1T parameter MoE model on 15.5T tokens with zero training instability.
- MuonClip Optimizer: We apply the Muon optimizer to an unprecedented scale, and develop novel optimization techniques to resolve instabilities while scaling up.
- Agentic Intelligence: Specifically designed for tool use, reasoning, and autonomous problem-solving.
Model Variants
- Kimi-K2-Base: The foundation model, a strong start for researchers and builders who want full control for fine-tuning and custom solutions.
- Kimi-K2-Instruct: The post-trained model best for drop-in, general-purpose chat and agentic experiences. It is a reflex-grade model without long thinking.
<picture>
</picture>
2. Model Summary
| Architecture | Mixture-of-Experts (MoE) |
| Total Parameters | 1T |
| Activated Parameters | 32B |
| Number of Layers (Dense layer included) | 61 |
| Number of Dense Layers | 1 |
| Attention Hidden Dimension | 7168 |
| MoE Hidden Dimension (per Expert) | 2048 |
| Number of Attention Heads | 64 |
| Number of Experts | 384 |
| Selected Experts per Token | 8 |
| Number of Shared Experts | 1 |
| Vocabulary Size | 160K |
| Context Length | 128K |
| Attention Mechanism | MLA |
| Activation Function | SwiGLU |
3. Evaluation Results
Instruction model evaluation results
| Benchmark | Metric | <sup>Kimi K2 Instruct</sup> | <sup>DeepSeek-V3-0324</sup> | <sup>Qwen3-235B-A22B <sup>(non-thinking)</sup></sup> |
<sup>Claude Sonnet 4 <sup>(w/o extended thinking)</sup></sup> |
<sup>Claude Opus 4 <sup>(w/o extended thinking)</sup></sup> |
<sup>GPT-4.1</sup> | <sup>Gemini 2.5 Flash Preview (05-20)</sup> |
|---|---|---|---|---|---|---|---|---|
| Coding Tasks | ||||||||
| LiveCodeBench v6 <sup>(Aug 24 - May 25)</sup> |
Pass@1 | 53.7 | 46.9 | 37.0 | 48.5 | 47.4 | 44.7 | 44.7 |
| OJBench | Pass@1 | 27.1 | 24.0 | 11.3 | 15.3 | 19.6 | 19.5 | 19.5 |
| MultiPL-E | Pass@1 | <ins>85.7</ins> | 83.1 | 78.2 | 88.6 | 89.6 | 86.7 | 85.6 |
| SWE-bench Verified <sup>(Agentless Coding)</sup> |
Single Patch w/o Test (Acc) | <ins>51.8</ins> | 36.6 | 39.4 | 50.2 | 53.0 | 40.8 | 32.6 |
| SWE-bench Verified <sup>(Agentic Coding)</sup> |
Single Attempt (Acc) | <ins>65.8</ins> | 38.8 | 34.4 | 72.7<sup>*</sup> | 72.5<sup>*</sup> | 54.6 | — |
| Multiple Attempts (Acc) | <ins>71.6</ins> | — | — | 80.2 | 79.4<sup>*</sup> | — | — | |
| SWE-bench Multilingual <sup>(Agentic Coding)</sup> |
Single Attempt (Acc) | <ins>47.3 </ins> | 25.8 | 20.9 | 51.0 | — | 31.5 | — |
| TerminalBench | Inhouse Framework (Acc) | <ins>30.0</ins> | — | — | 35.5 | 43.2 | 8.3 | — |
| Terminus (Acc) | <ins>25.0 </ins> | 16.3 | 6.6 | — | — | 30.3 | 16.8 | |
| Aider-Polyglot | Acc | 60.0 | 55.1 | <ins>61.8</ins> | 56.4 | 70.7 | 52.4 | 44.0 |
| Tool Use Tasks | ||||||||
| Tau2 retail | Avg@4 | <ins>70.6</ins> | 69.1 | 57.0 | 75.0 | 81.8 | 74.8 | 64.3 |
| Tau2 airline | Avg@4 | <ins>56.5</ins> | 39.0 | 26.5 | 55.5 | 60.0 | 54.5 | 42.5 |
| Tau2 telecom | Avg@4 | 65.8 | 32.5 | 22.1 | 45.2 | 57.0 | 38.6 | 16.9 |
| AceBench | Acc | <ins>76.5</ins> | 72.7 | 70.5 | 76.2 | 75.6 | 80.1 | 74.5 |
| Math & STEM Tasks | ||||||||
| AIME 2024 | Avg@64 | 69.6 | 59.4<sup>*</sup> | 40.1<sup>*</sup> | 43.4 | 48.2 | 46.5 | 61.3 |
| AIME 2025 | Avg@64 | 49.5 | 46.7 | 24.7<sup>*</sup> | 33.1<sup>*</sup> | 33.9<sup>*</sup> | 37.0 | 46.6 |
| MATH-500 | Acc | 97.4 | 94.0<sup>*</sup> | 91.2<sup>*</sup> | 94.0 | 94.4 | 92.4 | 95.4 |
| HMMT 2025 | Avg@32 | 38.8 | 27.5 | 11.9 | 15.9 | 15.9 | 19.4 | 34.7 |
| CNMO 2024 | Avg@16 | 74.3 | <ins>74.7</ins> | 48.6 | 60.4 | 57.6 | 56.6 | 75.0 |
| PolyMath-en | Avg@4 | 65.1 | 59.5 | 51.9 | 52.8 | 49.8 | 54.0 | 49.9 |
| ZebraLogic | Acc | 89.0 | 84.0 | 37.7<sup>*</sup> | 73.7 | 59.3 | 58.5 | 57.9 |
| AutoLogi | Acc | <ins>89.5</ins> | 88.9 | 83.3 | 89.8 | 86.1 | 88.2 | 84.1 |
| GPQA-Diamond | Avg@8 | 75.1 | 68.4<sup>*</sup> | 62.9<sup>*</sup> | 70.0<sup>*</sup> | 74.9<sup>*</sup> | 66.3 | 68.2 |
| SuperGPQA | Acc | 57.2 | 53.7 | 50.2 | 55.7 | 56.5 | 50.8 | 49.6 |
| Humanity's Last Exam <sup>(Text Only)</sup> |
- | 4.7 | 5.2 | <ins>5.7</ins> | 5.8 | 7.1 | 3.7 | 5.6 |
| General Tasks | ||||||||
| MMLU | EM | <ins>89.5</ins> | 89.4 | 87.0 | 91.5 | 92.9 | 90.4 | 90.1 |
| MMLU-Redux | EM | <ins>92.7</ins> | 90.5 | 89.2 | 93.6 | 94.2 | 92.4 | 90.6 |
| MMLU-Pro | EM | 81.1 | <ins>81.2</ins><sup>*</sup> | 77.3 | 83.7 | 86.6 | 81.8 | 79.4 |
| IFEval | Prompt Strict | 89.8 | 81.1 | 83.2<sup>*</sup> | 87.6 | 87.4 | 88.0 | 84.3 |
| Multi-Challenge | Acc | 54.1 | 31.4 | 34.0 | 46.8 | 49.0 | 36.4 | 39.5 |
| SimpleQA | Correct | <ins>31.0</ins> | 27.7 | 13.2 | 15.9 | 22.8 | 42.3 | 23.3 |
| Livebench | Pass@1 | 76.4 | 72.4 | 67.6 | 74.8 | 74.6 | 69.8 | 67.8 |
<sup>
• Bold denotes global SOTA, and underlined denotes open-source SOTA.
</sup>
<sup>
• Data points marked with * are taken directly from the model’s tech report or blog.
</sup>
<sup>
• All metrics, except for SWE-bench Verified (Agentless), are evaluated with an 8k output token length. SWE-bench Verified (Agentless) is limited to a 16k output token length.
</sup>
<sup>
• Kimi K2 achieves 65.8% pass@1 on the SWE-bench Verified tests with bash/editor tools (single-attempt patches, no test-time compute). It also achieves a 47.3% pass@1 on the SWE-bench Multilingual tests under the same conditions. Additionally, we report results on SWE-bench Verified tests (71.6%) that leverage parallel test-time compute by sampling multiple sequences and selecting the single best via an internal scoring model.
</sup>
<sup>
• To ensure the stability of the evaluation, we employed avg@k on the AIME, HMMT, CNMO, PolyMath-en, GPQA-Diamond, EvalPlus, Tau2.
</sup>
<sup>
• Some data points have been omitted due to prohibitively expensive evaluation costs.
</sup>
Base model evaluation results
<div align="center">| Benchmark | Metric | Shot | Kimi K2 Base | Deepseek-V3-Base | Qwen2.5-72B | Llama 4 Maverick |
|---|---|---|---|---|---|---|
| General Tasks | ||||||
| MMLU | EM | 5-shot | 87.8 | 87.1 | 86.1 | 84.9 |
| MMLU-pro | EM | 5-shot | 69.2 | 60.6 | 62.8 | 63.5 |
| MMLU-redux-2.0 | EM | 5-shot | 90.2 | 89.5 | 87.8 | 88.2 |
| SimpleQA | Correct | 5-shot | 35.3 | 26.5 | 10.3 | 23.7 |
| TriviaQA | EM | 5-shot | 85.1 | 84.1 | 76.0 | 79.3 |
| GPQA-Diamond | Avg@8 | 5-shot | 48.1 | 50.5 | 40.8 | 49.4 |
| SuperGPQA | EM | 5-shot | 44.7 | 39.2 | 34.2 | 38.8 |
| Coding Tasks | ||||||
| LiveCodeBench v6 | Pass@1 | 1-shot | 26.3 | 22.9 | 21.1 | 25.1 |
| EvalPlus | Pass@1 | - | 80.3 | 65.6 | 66.0 | 65.5 |
| Mathematics Tasks | ||||||
| MATH | EM | 4-shot | 70.2 | 60.1 | 61.0 | 63.0 |
| GSM8k | EM | 8-shot | 92.1 | 91.7 | 90.4 | 86.3 |
| Chinese Tasks | ||||||
| C-Eval | EM | 5-shot | 92.5 | 90.0 | 90.9 | 80.9 |
| CSimpleQA | Correct | 5-shot | 77.6 | 72.1 | 50.5 | 53.5 |
• All models are evaluated using the same evaluation protocol. ## 5. Model Usage ### Chat Completion Once the local inference service is up, you can interact with it through the chat endpoint: ### Tool Calling Kimi-K2-Instruct has strong tool-calling capabilities. To enable them, you need to pass the list of available tools in each request, then the model will autonomously decide when and how to invoke them. ## 6. License Both the code repository and the model weights are released under the [Modified MIT License](LICENSE). --- ## 7. Third Party Notices See [THIRD PARTY NOTICES](THIRD_PARTY_NOTICES.md) --- ## 7. Contact Us If you have any questions, please reach out at [support@moonshot.cn](mailto:support@moonshot.cn).