Moonshot AI
availableShows if the model has enough results for an index.Kimi K2.6
Kimi K2.6 is a reasoning model from Moonshot AI. 34 benchmarks count toward its score, in 7 categories.
IndexOverall score. 50 is the middle.60.9 ±3.3
CoverageShare of the index weight with results.95%
SpeedOutput tokens per second.38/s
Input / 1MUS dollars per 1M input tokens.$0.95
Output / 1MUS dollars per 1M output tokens.$4
ContextMaximum tokens in one request.256K
EloLMArena rating and rank.N/A
50 is the middle of the board. The range shows the doubt in the index.
CapabilitiesScore per category. 50 is the middle.
50 is the middleResults
34 counted| BenchmarkThe test name. | CategoryThe capability that the test measures. | ResultThe score from the publisher. | IndexThis result on the index scale. | RunThe settings of the run. | DateDate of the result. | Published byThe source of the result. |
|---|---|---|---|---|---|---|
| V* | Multimodal | 96.9% | 57.6 | — | — | Z.AI |
| AIME 2026 | Math | 96.4% | 55.3 | — | — | Qwen |
| τ²-Bench Tool-Agent-User Evaluation | Agentic | 95.9% | 67.6 | — | — | Victor Barres et al. |
| Harvard-MIT Mathematics Tournament February 2026 | Math | 92.7% | 59.5 | — | — | Qwen |
| DeepSearchQA | Agentic | 92.5% | 71.6 | — | — | Meta AI |
| Artificial Analysis GPQA Diamond | Knowledge | 91.1% | 61.9 | — | — | Artificial Analysis |
| Graduate-Level Google-Proof Q&A | Knowledge | 90.5% | 61.8 | — | — | David Rein et al. |
| GPQA Diamond | Knowledge | 90.5% | 61.8 | — | — | David Rein et al. |
| LiveCodeBench v6 | Coding | 89.6% | 58.6 | — | — | LiveCodeBench maintainers |
| MathVision | Multimodal | 87.4% | — | — | — | Qwen |
| MMAnswerBench | Math | 86.0% | — | — | — | Qwen |
| BrowseComp | Agentic | 83.2% | 69.6 | — | — | OpenAI |
| Artificial Analysis Long Context Reasoning | Reasoning | 81.0% | 64.3 | — | — | Artificial Analysis |
| WideResearch | Agentic | 80.8% | 66.7 | — | — | Qwen |
| CharXiv Reasoning | Multimodal | 80.4% | 57.6 | — | — | CharXiv authors |
| Software Engineering Benchmark Verified | Coding | 80.2% | 61.9 | — | — | Carlos E. Jimenez et al. |
| MMMU-Pro with Python | Multimodal | 80.1% | — | — | — | OpenAI |
| Massive Multi-discipline Multimodal Understanding Pro | Multimodal | 79.4% | 56.6 | — | — | MMMU-Pro authors |
| Artificial Analysis MMMU-Pro | Multimodal | 79.4% | 64.1 | — | — | Artificial Analysis |
| Artificial Analysis IFBench | Instruction | 76.0% | 66.2 | — | — | Artificial Analysis |
| OSWorld-Verified | Agentic | 73.1% | 61.8 | — | — | Tianbao Xie et al. |
| Claw-Eval | Agentic | 62.3% | 56.3 | — | — | Bowen Ye et al. |
| Artificial Analysis Coding Index | Coding | 61.8% | 62.5 | — | — | Artificial Analysis |
| SWE-bench Pro | Coding | 58.6% | 60.7 | — | — | Xiang Deng et al. |
| Gert Labs Composite Game Benchmark | Agentic | 56.8% | 62.9 | — | — | Gert Labs |
| MCP Atlas | Agentic | 55.9% | 50.3 | — | — | OpenAI |
| Scientific Code Benchmark | Coding | 52.2% | 62.9 | — | — | Benchmark authors |
| Artificial Analysis SciCode | Coding | 51.5% | 64.3 | — | — | Artificial Analysis |
| Toolathlon | Agentic | 50.0% | 64.6 | — | — | OpenAI |
| cursorBench31 | Coding | 47.6% | — | — | — | Benchmark authors |
| Artificial Analysis Humanity's Last Exam | Knowledge | 37.5% | 65.2 | — | — | Artificial Analysis |
| Humanity's Last Exam | Knowledge | 34.7% | 58.2 | — | — | Center for AI Safety et al. |
| Artificial Analysis Omniscience Accuracy | Knowledge | 32.6% | 54.2 | — | — | Artificial Analysis |
| APEX-Agents-AA | Agentic | 28.5% | 63.6 | — | — | Artificial Analysis / Mercor |
| Artificial Analysis Intelligence Index | Knowledge | 27.0% | 56.1 | — | — | Artificial Analysis |
| GDPval-AA normalized | Agentic | 26.3% | 55.8 | — | — | Artificial Analysis |
| Artificial Analysis Agentic Index | Agentic | 22.1% | 55.2 | — | — | Artificial Analysis |
| ResearchClawBench | Agentic | 18.0% | — | — | — | InternScience |
| Critical Physics Tasks | Reasoning | 8.0% | 55.2 | — | — | Artificial Analysis |
| OSWorld 2.0 | Agentic | 4.6% | 58.2 | — | — | Mengqi Yuan et al. |
34 benchmarks count, from 35 of 40 results. A grey row does not count. Too few models took that benchmark.