Moonshot · model
Kimi K2.5 benchmark scores
As of 2026-10-07, Kimi K2.5 (Moonshot) has measured scores on 54 benchmarks and estimated scores on 77 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 87.9% | measured |
| AA-HLE | 30.7% | measured |
| AA-SciCode | 48.3% | estimated ± 3.1 pp, high confidence |
| AIME26 | 95.8% | measured |
| Artificial Analysis Intelligence Index | 23.5% | measured |
| CritPt | 3.1% | measured |
| GDPval-AA | 17.2% | measured |
| GPQA Diamond | 87.6% | measured |
| HLE | 30.1% | measured |
| IFBench | 68.8% | estimated ± 6.9 pp, medium confidence |
| AA-LCR | 78.0% | measured |
| LiveCodeBench v6 | 85.0% | measured |
| AA-Omniscience Accuracy | 35.2% | measured |
| SciCode | 48.7% | measured |
| BrowseComp | 60.6% | measured |
| HLE w/ tools | 41.3% | estimated ± 2.4 pp, medium confidence |
| IFEval | 93.9% | measured |
| LongBench v2 | 61.0% | measured |
| VITA-Bench | 38.9% | estimated ± 8.2 pp, low confidence |
| τ²-bench results | 95.9% | measured |
| AA Coding Index | 46.8% | measured |
| AA-MMMU-Pro | 75.4% | measured |
| APEX-Agents | 0.0% | estimated ± 0.6 pp, low confidence |
| APEX-Agents-AA | 11.5% | measured |
| BioMysteryBench (human-difficult) | 38.3% | estimated ± 5.9 pp, low confidence |
| DeepSearchQA | 77.1% | measured |
| SWE-bench Verified | 76.8% | measured |
| AutomationBench | 28.4% | estimated ± 9.7 pp, low confidence |
| BFCL v4 | 71.8% | estimated ± 2.6 pp, high confidence |
| CyberGym | 64.5% | estimated ± 4.6 pp, medium confidence |
| JobBench | 8.7% | measured |
| SkillsBench | 42.8% | estimated ± 5.6 pp, low confidence |
| SWE-bench Pro | 50.7% | measured |
| τ³-bench results | 65.7% | measured |
| WideResearch | 72.7% | measured |
| AA Agentic Index | 11.6% | estimated ± 2.8 pp, high confidence |
| MMLU-Pro | 87.1% | measured |
| FrontierMath v2 (Tier 4) | 4.2% | measured |
| FrontierMath v2 (Tiers 1-3) | 27.9% | measured |
| GPQA | 87.6% | measured |
| AA-IFBench | 70.2% | measured |
| Gert Labs | 45.9% | measured |
| AA-AnalystAgent | 11.5% | estimated ± 12.0 pp, low confidence |
| AA EnterpriseOps-Gym | 48.1% | estimated ± 3.6 pp, medium confidence |
| AA Harvey LAB | 85.6% | estimated ± 9.8 pp, low confidence |
| ApprenticeBench | 0.0% | estimated ± 3.1 pp, low confidence |
| ARC-AGI-1 | 85.2% | estimated ± 2.7 pp, medium confidence |
| ARC-AGI-2 | 47.0% | estimated ± 10.1 pp, low confidence |
| CursorBench 3.1 | 47.9% | estimated ± 4.8 pp, low confidence |
| CursorBench 3.2 | 41.8% | estimated ± 3.6 pp, medium confidence |
| FrontierCode 1.1 Main | 15.2% | estimated ± 2.7 pp, low confidence |
| FrontierSWE v2 | 0.0% | estimated ± 9.0 pp, low confidence |
| OfficeQA Pro | 50.3% | estimated ± 7.2 pp, low confidence |
| OSWorld-Verified | 52.8% | estimated ± 5.4 pp, low confidence |
| Vals GPQA Diamond | 86.8% | estimated ± 2.0 pp, high confidence |
| Vals LiveCodeBench | 86.0% | estimated ± 0.8 pp, medium confidence |
| Vals MMLU-Pro | 85.0% | estimated ± 1.2 pp, high confidence |
| Vals SWE-bench | 77.5% | estimated ± 1.4 pp, medium confidence |
| Terminal-Bench 2.1 (Vals AI) | 44.4% | estimated ± 6.3 pp, medium confidence |
| AA AutomationBench | 42.0% | estimated ± 5.0 pp, low confidence |
| GDP.pdf | 17.8% | estimated ± 5.2 pp, low confidence |
| AA ITBench | 12.4% | estimated ± 3.6 pp, low confidence |
| AA Tau3 Banking | 22.0% | estimated ± 5.8 pp, medium confidence |
| CursorBench 4.0 | 14.3% | estimated ± 1.6 pp, medium confidence |
| CWE-bench v1 | 26.1% | estimated ± 6.9 pp, low confidence |
| DeepSWE | 9.9% | estimated ± 6.4 pp, low confidence |
| HLE w/o tools | 25.9% | estimated ± 2.9 pp, high confidence |
| OSWorld 2.0 | 0.0% | estimated ± 4.5 pp, low confidence |
| PostTrainBench v1.1 | 20.3% | estimated ± 4.7 pp, medium confidence |
| SWE Multilingual | 73.0% | measured |
| Toolathlon-Verified | 73.2% | estimated ± 1.3 pp, low confidence |
| Vibe Code Bench | 27.9% | estimated ± 6.5 pp, low confidence |
| CharXiv | 81.6% | estimated ± 5.4 pp, medium confidence |
| CharXiv w/o tools | 82.8% | estimated ± 1.8 pp, medium confidence |
| ExploitGym | 0.0% | estimated ± 6.1 pp, low confidence |
| C-Eval | 92.2% | estimated ± 0.7 pp, medium confidence |
| Claw-Eval | 52.3% | measured |
| DeepPlanning | 14.4% | measured |
| HMMT Feb 2025 | 95.4% | measured |
| HMMT Feb 2026 | 87.1% | measured |
| HMMT Nov 2025 | 91.1% | measured |
| MathVision | 90.3% | estimated ± 3.2 pp, medium confidence |
| MCP Atlas | 29.5% | measured |
| MCP-Tasks | 59.1% | measured |
| MMAnswerBench | 81.8% | measured |
| MMLU-ProX | 82.3% | measured |
| MMLU-Redux | 94.2% | estimated ± 0.7 pp, high confidence |
| MMMU-Pro | 78.5% | measured |
| NL2Repo | 44.4% | estimated ± 3.1 pp, low confidence |
| NOVA-63 | 56.0% | measured |
| QwenClawBench | 54.3% | measured |
| ScreenSpot Pro | 74.2% | estimated ± 10.0 pp, low confidence |
| SuperGPQA | 69.2% | measured |
| Terminal-Bench 2.0 | 50.8% | measured |
| Toolathlon | 27.8% | measured |
| VideoMMMU | 86.6% | measured |
| ERQA | 64.0% | estimated ± 3.9 pp, high confidence |
| HealthBench Hard | 0.0% | estimated ± 5.1 pp, low confidence |
| LiveCodeBench Pro | 76.0% | estimated ± 6.0 pp, low confidence |
| MedXpertQA (MM) | 71.3% | estimated ± 2.4 pp, high confidence |
| MedXpertQA (Text) | 47.4% | estimated ± 4.5 pp, low confidence |
| MMLU-Pro (Arcee) | 87.1% | measured |
| React Native Evals | 77.2% | measured |
| ResearchClawBench | 14.0% | measured |
| SWE-Rebench | 58.5% | measured |
| ARC-AGI-3 | 0.0% | estimated ± 2.1 pp, medium confidence |
| FrontierMath (legacy) | 28.7% | estimated ± 0.7 pp, low confidence |
| BioMysteryBench (human-solvable) | 81.7% | estimated ± 0.6 pp, low confidence |
| DRACO | 68.6% | estimated ± 3.0 pp, low confidence |
| FrontierCode 1.1 Extended | 52.6% | estimated ± 3.0 pp, low confidence |
| HealthBench (raw) | 39.9% | estimated ± 6.4 pp, low confidence |
| HealthBench (length-adjusted) | 53.9% | estimated ± 3.7 pp, low confidence |
| HealthBench Professional | 45.5% | estimated ± 4.8 pp, low confidence |
| HealthBench Professional (raw) | 51.0% | estimated ± 5.0 pp, low confidence |
| HLE-Verified | 15.4% | estimated ± 2.0 pp, low confidence |
| LABBench2 | 73.8% | estimated ± 1.8 pp, low confidence |
| AIME 2025 | 96.1% | measured |
| MMMU | 83.3% | estimated ± 0.9 pp, high confidence |
| LiveCodeBench | 90.5% | estimated ± 8.9 pp, low confidence |
| Agents' Last Exam | 48.9% | estimated ± 8.7 pp, low confidence |
| OpenHarmony Bench | 46.0% | estimated ± 2.8 pp, medium confidence |
| IMOAnswerBench | 82.0% | estimated ± 5.9 pp, medium confidence |
| MMVU | 80.4% | measured |
| SimpleVQA | 67.3% | estimated ± 8.2 pp, medium confidence |
| MMMLU | 88.9% | estimated ± 1.1 pp, medium confidence |
| MMLU | 92.3% | estimated ± 1.1 pp, low confidence |
| MMMU-Pro w/ Python | 79.8% | estimated ± 0.6 pp, high confidence |
| MATH-500 | 97.7% | estimated ± 3.1 pp, medium confidence |
| RealWorldQA | 85.2% | estimated ± 1.2 pp, medium confidence |
| Video-MME (with subtitle) | 88.6% | estimated ± 1.2 pp, medium confidence |
| OmniDocBench 1.5 | 91.7% | estimated ± 5.9 pp, low confidence |