Moonshot · model
Kimi K3.1 Pro benchmark scores
As of 2026-10-29, Kimi K3.1 Pro (Moonshot) has measured scores on 39 benchmarks and estimated scores on 103 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 92.4% | estimated ± 2.7 pp, high confidence |
| AA-HLE | 52.5% | measured |
| AA-SciCode | 60.8% | measured |
| CritPt | 24.6% | measured |
| GDPval-AA | 67.3% | measured |
| GPQA Diamond | 94.2% | measured |
| HLE | 70.0% | measured |
| AA-LCR | 86.3% | measured |
| LiveCodeBench v6 | 100.0% | estimated ± 4.9 pp, medium confidence |
| SciCode | 60.3% | estimated ± 3.1 pp, medium confidence |
| Terminal-Bench 2.1 | 87.4% | estimated ± 5.2 pp, low confidence |
| BrowseComp | 92.1% | measured |
| HLE w/ tools | 70.0% | measured |
| LongBench v2 | 63.3% | estimated ± 5.8 pp, low confidence |
| VITA-Bench | 47.4% | estimated ± 6.7 pp, low confidence |
| τ²-bench results | 98.9% | estimated ± 3.1 pp, medium confidence |
| AA Coding Index | 77.9% | estimated ± 2.5 pp, medium confidence |
| AA-MMMU-Pro | 83.9% | measured |
| APEX-Agents | 67.2% | measured |
| APEX-Agents-AA | 39.0% | estimated ± 2.9 pp, low confidence |
| BioMysteryBench (human-difficult) | 49.0% | estimated ± 6.8 pp, low confidence |
| DeepSearchQA | 93.9% | estimated ± 7.3 pp, low confidence |
| SWE-bench Verified | 94.4% | estimated ± 4.6 pp, medium confidence |
| AutomationBench | 47.6% | estimated ± 9.5 pp, low confidence |
| CyberGym | 96.3% | estimated ± 5.9 pp, low confidence |
| JobBench | 66.5% | estimated ± 4.9 pp, low confidence |
| SWE-bench Pro | 76.1% | estimated ± 6.6 pp, medium confidence |
| WideResearch | 83.1% | estimated ± 3.5 pp, high confidence |
| AA Agentic Index | 68.0% | estimated ± 2.9 pp, medium confidence |
| MMLU-Pro | 85.9% | estimated ± 2.5 pp, low confidence |
| Artificial Analysis Intelligence Index | 46.0% | measured |
| GPQA | 94.2% | measured |
| MMLU | 95.0% | estimated ± 1.0 pp, medium confidence |
| AA-Omniscience Accuracy | 48.9% | measured |
| MMMLU | 91.2% | estimated ± 0.7 pp, low confidence |
| MMMU | 88.5% | estimated ± 0.8 pp, medium confidence |
| AA MMLU-Pro | 90.3% | estimated ± 0.6 pp, medium confidence |
| Gert Labs | 79.2% | estimated ± 7.4 pp, low confidence |
| AA-AnalystAgent | 38.8% | measured |
| AA EnterpriseOps-Gym | 46.5% | estimated ± 3.3 pp, medium confidence |
| AA Harvey LAB v1.0 | 94.0% | estimated ± 1.3 pp, low confidence |
| AA Tau3 Banking | 45.2% | estimated ± 3.9 pp, high confidence |
| AA Terminal-Bench 2.1 | 87.5% | estimated ± 4.8 pp, low confidence |
| ApprenticeBench | 17.6% | estimated ± 12.5 pp, low confidence |
| ARC-AGI-1 | 94.9% | estimated ± 2.7 pp, high confidence |
| ARC-AGI-2 | 79.9% | estimated ± 10.1 pp, low confidence |
| CursorBench 3.1 | 70.6% | estimated ± 3.4 pp, medium confidence |
| CursorBench 3.2 | 70.9% | estimated ± 5.3 pp, medium confidence |
| FrontierCode 1.1 Main | 55.0% | estimated ± 3.7 pp, low confidence |
| FrontierSWE v2 | 43.5% | estimated ± 11.8 pp, low confidence |
| MazeBench | 2.0% | measured |
| OfficeQA Pro | 67.4% | measured |
| OSWorld-Verified | 86.7% | estimated ± 5.8 pp, low confidence |
| ParseBench | 71.6% | estimated ± 0.8 pp, low confidence |
| Vals GPQA Diamond | 93.6% | estimated ± 2.0 pp, high confidence |
| Vals MMLU-Pro | 88.5% | estimated ± 1.2 pp, high confidence |
| Vals SWE-bench | 92.6% | estimated ± 7.5 pp, low confidence |
| Terminal-Bench 2.1 (Vals AI) | 93.5% | estimated ± 5.4 pp, low confidence |
| VulcanBench v3 | 87.6% | estimated ± 3.9 pp, medium confidence |
| AA AutomationBench | 73.3% | measured |
| CWE-Bench-AA | 60.8% | measured |
| AA Cyber Index | 64.1% | measured |
| DeepsecBench-AA | 36.7% | measured |
| GDP.pdf | 31.4% | measured |
| AA ITBench | 50.9% | estimated ± 4.1 pp, high confidence |
| AA Terminal-Bench 4.0 | 67.2% | measured |
| Bug Hunt Bench | 36.0% | measured |
| CursorBench 4.0 | 49.8% | estimated ± 3.5 pp, high confidence |
| CWE-bench v1 | 67.8% | estimated ± 2.7 pp, medium confidence |
| DeepSWE | 75.5% | measured |
| FrontierCode 2.0 | 30.4% | estimated ± 7.0 pp, low confidence |
| GraphWalks BFS 256K–1M | 71.9% | estimated ± 14.8 pp, low confidence |
| HLE w/o tools | 51.1% | measured |
| OSWorld 2.0 | 79.6% | measured |
| OSWorld 2.1 | 79.6% | measured |
| PostTrainBench v1.1 | 43.9% | estimated ± 4.5 pp, high confidence |
| Terminal-Bench 4.0 | 65.7% | measured |
| Toolathlon-Verified | 80.2% | estimated ± 1.0 pp, low confidence |
| OSWorld 2.1 Offline | 87.1% | estimated ± 7.9 pp, low confidence |
| ARC-AGI-3 | 3.5% | estimated ± 2.0 pp, high confidence |
| ARC-AGI-3 Provider Adapter | 94.3% | estimated ± 7.5 pp, medium confidence |
| Hermes Bench | 73.3% | estimated ± 0.3 pp, medium confidence |
| Hermes Index | 44.9% | estimated ± 2.6 pp, medium confidence |
| CharXiv | 93.4% | measured |
| CharXiv w/o tools | 89.0% | measured |
| ExploitBench | 74.7% | measured |
| ExploitGym | 20.7% | measured |
| AI2D_TEST | 93.5% | estimated ± 0.9 pp, low confidence |
| C-Eval | 93.5% | estimated ± 1.2 pp, medium confidence |
| Claw-Eval | 81.7% | estimated ± 7.6 pp, low confidence |
| CountBench | 100.0% | estimated ± 2.9 pp, low confidence |
| DynaMath | 88.5% | estimated ± 0.8 pp, low confidence |
| ERQA | 79.1% | estimated ± 2.9 pp, low confidence |
| MathVision | 93.7% | estimated ± 2.9 pp, medium confidence |
| MCP Atlas | 86.0% | estimated ± 3.8 pp, low confidence |
| MLVU (M-Avg) | 90.1% | estimated ± 1.5 pp, high confidence |
| MMLU-Redux | 97.1% | estimated ± 0.8 pp, low confidence |
| MMMU-Pro | 84.1% | measured |
| NL2Repo | 62.3% | estimated ± 6.2 pp, medium confidence |
| RealWorldQA | 88.5% | estimated ± 1.5 pp, high confidence |
| ScreenSpot Pro | 89.7% | estimated ± 7.1 pp, medium confidence |
| SimpleVQA | 73.5% | estimated ± 7.5 pp, low confidence |
| SuperGPQA | 100.0% | estimated ± 6.9 pp, low confidence |
| Toolathlon | 63.4% | estimated ± 4.2 pp, high confidence |
| V* | 100.0% | estimated ± 7.5 pp, low confidence |
| Video-MME (w/o subtitle) | 87.5% | estimated ± 1.4 pp, low confidence |
| Video-MME (with subtitle) | 92.0% | estimated ± 2.9 pp, high confidence |
| VideoMMMU | 88.1% | estimated ± 1.5 pp, medium confidence |
| We-Math | 84.9% | estimated ± 7.4 pp, low confidence |
| HealthBench Hard | 36.5% | estimated ± 5.1 pp, low confidence |
| MedXpertQA (MM) | 82.7% | estimated ± 2.4 pp, medium confidence |
| MedXpertQA (Text) | 68.9% | estimated ± 4.5 pp, medium confidence |
| MMLU-Pro (Arcee) | 88.2% | estimated ± 3.4 pp, low confidence |
| React Native Evals | 81.3% | estimated ± 4.7 pp, low confidence |
| ResearchClawBench | 20.5% | estimated ± 2.2 pp, low confidence |
| Finance Agent v2 | 56.9% | estimated ± 3.8 pp, medium confidence |
| BioMysteryBench (human-solvable) | 89.4% | estimated ± 0.6 pp, medium confidence |
| Chartography (tools) | 90.2% | measured |
| DRACO | 89.2% | estimated ± 3.1 pp, low confidence |
| HealthBench (raw) | 59.4% | estimated ± 6.0 pp, medium confidence |
| HealthBench (length-adjusted) | 53.9% | estimated ± 3.3 pp, high confidence |
| HealthBench Professional | 63.1% | estimated ± 4.5 pp, medium confidence |
| HealthBench Professional (raw) | 69.5% | estimated ± 4.5 pp, high confidence |
| HLE-Verified | 54.4% | estimated ± 6.8 pp, low confidence |
| LABBench2 | 85.1% | estimated ± 2.4 pp, medium confidence |
| STEMBench | 2.4% | estimated ± 0.2 pp, high confidence |
| SimpleQA | 52.5% | estimated ± 14.6 pp, low confidence |
| Agents' Last Exam | 43.2% | measured |
| OpenHarmony Bench | 58.9% | estimated ± 2.4 pp, medium confidence |
| SEC-Bench Pro | 75.6% | estimated ± 3.7 pp, medium confidence |
| BabyVision | 70.4% | estimated ± 11.4 pp, low confidence |
| MMVU | 80.8% | estimated ± 2.7 pp, low confidence |
| ZeroBench | 39.0% | measured |
| MRCRv2 | 86.0% | estimated ± 11.3 pp, low confidence |
| LVBench | 86.8% | estimated ± 8.2 pp, low confidence |
| MMLongBench-Doc | 61.7% | estimated ± 2.6 pp, low confidence |
| SWE-Marathon | 51.7% | measured |
| MMMU-Pro w/ Python | 90.0% | measured |
| OCRBench V2 | 74.1% | estimated ± 4.4 pp, low confidence |
| MathVision w/ Python | 97.7% | estimated ± 0.9 pp, medium confidence |
| PinchBench | 90.8% | estimated ± 3.6 pp, low confidence |
| SWE-Atlas Codebase QnA | 64.1% | estimated ± 5.4 pp, low confidence |