Moonshot AI
availableShows if the model has enough results for an index.Kimi K3
Kimi K3 is a reasoning model from Moonshot AI. 38 benchmarks count toward its score, in 5 categories.
IndexOverall score. 50 is the middle.72.0 ±4.6
CoverageShare of the index weight with results.80%
SpeedOutput tokens per second.37/s
Input / 1MUS dollars per 1M input tokens.$3
Output / 1MUS dollars per 1M output tokens.$15
ContextMaximum tokens in one request.1.05M
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
38 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. |
|---|---|---|---|---|---|---|
| MathVision with Python | Multimodal | 97.8% | — | — | — | Moonshot AI / MathVision authors |
| DeepSearchQA | Agentic | 95.0% | 73.6 | — | — | Meta AI |
| Artificial Analysis Harvey LAB-AA | Agentic | 94.6% | 79.0 | — | — | Artificial Analysis |
| MathVision | Multimodal | 94.3% | — | — | — | Qwen |
| Graduate-Level Google-Proof Q&A | Knowledge | 93.5% | 64.6 | — | — | David Rein et al. |
| GPQA Diamond | Knowledge | 93.5% | 64.6 | — | — | David Rein et al. |
| Artificial Analysis GPQA Diamond | Knowledge | 93.5% | 64.4 | — | — | Artificial Analysis |
| CharXiv Reasoning | Multimodal | 91.3% | 70.1 | — | — | CharXiv authors |
| BrowseComp | Agentic | 91.2% | 76.3 | — | — | OpenAI |
| OmniDocBench | Multimodal | 91.1% | — | — | — | Moonshot AI / OmniDocBench authors |
| Artificial Analysis Long Context Reasoning | Reasoning | 88.7% | 69.6 | — | — | Artificial Analysis |
| BabyVision with Python | Multimodal | 85.7% | — | — | — | Moonshot AI |
| CharXiv Reasoning without tools | Multimodal | 84.8% | — | — | — | CharXiv authors |
| MCP Atlas | Agentic | 84.2% | 71.5 | — | — | OpenAI |
| MMMU-Pro with Python | Multimodal | 83.4% | — | — | — | OpenAI |
| Massive Multi-discipline Multimodal Understanding Pro | Multimodal | 81.6% | 60.1 | — | — | MMMU-Pro authors |
| FrontierSWE | Coding | 81.2% | — | — | — | Evan Chu et al. |
| Artificial Analysis MMMU-Pro | Multimodal | 80.5% | 65.5 | — | — | Artificial Analysis |
| ProgramBench: Can Language Models Rebuild Programs From Scratch? | Coding | 77.8% | 75.3 | — | — | John Yang et al. |
| Artificial Analysis Coding Index | Coding | 76.2% | 72.7 | — | — | Artificial Analysis |
| VulcanBench v3 | Coding | 73.7% | 51.2 | — | — | VulcanBench contributors |
| DECK-Bench (Internal) | Agentic | 73.5% | — | — | — | Moonshot AI |
| Toolathlon-Verified | Agentic | 73.2% | 71.5 | — | — | Moonshot AI |
| Kimi Code Bench v2 | Coding | 72.9% | — | — | — | Moonshot AI |
| DeepSWE | Agentic | 67.5% | 72.8 | — | — | Datacurve AI |
| OfficeQA Pro | Multimodal | 63.3% | 75.6 | — | — | OfficeQA Pro authors |
| cursorBench32 | Coding | 60.8% | 69.3 | — | — | Benchmark authors |
| Artificial Analysis SciCode | Coding | 59.5% | 75.4 | — | — | Artificial Analysis |
| PerceptionBench (Internal) | Multimodal | 58.5% | — | — | — | Moonshot AI |
| Artificial Analysis AutomationBench | Agentic | 58.3% | 69.8 | — | — | Artificial Analysis |
| OpenHarmony Bench v1.0 | Coding | 57.3% | 66.4 | — | — | OpenHarmony Bench authors |
| Humanity's Last Exam | Knowledge | 56.0% | 76.2 | — | — | Center for AI Safety et al. |
| JobBench | Agentic | 52.9% | 70.5 | — | — | Yuetai Li et al. |
| GDPval-AA normalized | Agentic | 51.2% | 74.9 | — | — | Artificial Analysis |
| WorldVQA ForceAnswer | Multimodal | 51.0% | — | — | — | Moonshot AI / WorldVQA authors |
| Artificial Analysis Agentic Index | Agentic | 50.6% | 78.5 | — | — | Artificial Analysis |
| MLS-Bench Lite | Coding | 48.3% | — | — | — | MLS-Bench |
| Artificial Analysis ITBench-AA | Agentic | 47.7% | — | — | — | Artificial Analysis |
| Artificial Analysis Omniscience Accuracy | Knowledge | 47.6% | 72.8 | — | — | Artificial Analysis |
| Artificial Analysis Humanity's Last Exam | Knowledge | 46.9% | 75.4 | — | — | Artificial Analysis |
| Artificial Analysis Tau3-Banking | Agentic | 46.0% | 75.1 | — | — | Artificial Analysis |
| Artificial Analysis EnterpriseOps-Gym | Agentic | 45.3% | 69.4 | — | — | Artificial Analysis |
| Artificial Analysis Intelligence Index | Knowledge | 43.6% | 76.9 | — | — | Artificial Analysis |
| Humanity's Last Exam without tools | Knowledge | 43.5% | 65.6 | — | — | OpenAI |
| SWE-Marathon | Coding | 42.0% | — | — | — | Abundant AI and BenchFlow |
| APEX-Agents-AA | Agentic | 41.3% | 73.7 | — | — | Artificial Analysis / Mercor |
| ZeroBench_main with Python | Multimodal | 41.0% | — | — | — | Moonshot AI / ZeroBench authors |
| Artificial Analysis AnalystAgent | Agentic | 38.8% | 68.7 | — | — | Artificial Analysis |
| Medical Long Context Reasoning (MLCR-AA) | Reasoning | 38.3% | 71.2 | — | — | Wisedocs and Artificial Analysis |
| APEX-Agents | Agentic | 37.6% | 70.7 | — | — | Moonshot AI / APEX-Agents benchmark authors |
| PostTrain Bench | Coding | 36.6% | — | — | — | Moonshot AI |
| SpreadsheetBench 2 | Agentic | 34.8% | — | — | — | Moonshot AI |
| AutomationBench | Agentic | 30.8% | 66.7 | — | — | Moonshot AI |
| FrontierSWE v2 | Coding | 25.9% | 68.8 | — | — | Proximal |
| Critical Physics Tasks | Reasoning | 23.4% | 87.4 | — | — | Artificial Analysis |
| ZeroBench | Multimodal | 23.0% | — | — | — | Meta AI |
| Artificial Analysis GDP.pdf | Agentic | 22.0% | 73.2 | — | — | Artificial Analysis |
| ApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable job | Agentic | 18.0% | 70.1 | — | — | NeoCognition |
38 benchmarks count, from 40 of 58 results. A grey row does not count. Too few models took that benchmark.