MiniMax · model
MiniMax M3 benchmark scores
As of 2026-11-21, MiniMax M3 (MiniMax) has measured scores on 45 benchmarks and estimated scores on 145 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 92.9% | measured |
| AA-HLE | 39.0% | measured |
| AA-SciCode | 47.1% | measured |
| Artificial Analysis Intelligence Index | 18.9% | measured |
| CritPt | 3.7% | measured |
| GDPval-AA | 37.1% | measured |
| GPQA Diamond | 92.3% | estimated ± 1.8 pp, high confidence |
| HLE | 43.4% | estimated ± 8.5 pp, medium confidence |
| IFBench | 82.3% | estimated ± 6.6 pp, low confidence |
| AA-LCR | 83.0% | measured |
| LiveCodeBench v6 | 89.1% | estimated ± 1.7 pp, high confidence |
| AA-Omniscience Accuracy | 44.9% | measured |
| SciCode | 41.1% | estimated ± 2.9 pp, low confidence |
| Terminal-Bench 2.1 | 66.0% | measured |
| BrowseComp | 83.5% | measured |
| HLE w/ tools | 51.7% | estimated ± 5.3 pp, medium confidence |
| IFEval | 93.3% | estimated ± 2.8 pp, medium confidence |
| LongBench v2 | 68.4% | estimated ± 5.5 pp, medium confidence |
| VITA-Bench | 45.8% | estimated ± 6.7 pp, low confidence |
| τ²-bench results | 88.9% | measured |
| AA Coding Index | 58.6% | measured |
| AA-MMMU-Pro | 78.6% | measured |
| APEX-Agents | 29.0% | estimated ± 9.4 pp, low confidence |
| APEX-Agents-AA | 28.2% | estimated ± 2.9 pp, medium confidence |
| BioMysteryBench (human-difficult) | 33.6% | estimated ± 6.8 pp, low confidence |
| DeepSearchQA | 89.4% | estimated ± 4.5 pp, high confidence |
| SWE-bench Verified | 80.5% | measured |
| AutomationBench | 47.6% | estimated ± 9.5 pp, low confidence |
| BFCL v4 | 75.4% | estimated ± 2.7 pp, medium confidence |
| CyberGym | 69.2% | estimated ± 5.3 pp, medium confidence |
| JobBench | 27.7% | estimated ± 4.9 pp, low confidence |
| SkillsBench | 55.3% | estimated ± 4.1 pp, medium confidence |
| SWE-bench Pro | 59.0% | measured |
| WideResearch | 81.3% | estimated ± 2.1 pp, medium confidence |
| AA Agentic Index | 30.8% | measured |
| MMLU-Pro | 84.4% | estimated ± 4.8 pp, high confidence |
| AA-IFBench | 82.9% | measured |
| FailureSensorIQ | 66.2% | estimated ± 1.7 pp, high confidence |
| GPQA | 93.9% | estimated ± 4.5 pp, high confidence |
| MMLU | 90.5% | estimated ± 0.9 pp, medium confidence |
| Vals GPQA Diamond | 92.7% | measured |
| Vals LiveCodeBench | 82.2% | measured |
| Vals MMLU-Pro | 84.2% | measured |
| MMMLU | 87.9% | estimated ± 0.7 pp, medium confidence |
| MMMU | 85.2% | estimated ± 0.8 pp, high confidence |
| Chess Puzzles | 17.0% | measured |
| AA MMLU-Pro | 88.1% | estimated ± 0.7 pp, medium confidence |
| Gert Labs | 56.9% | estimated ± 4.6 pp, medium confidence |
| AA-AnalystAgent | 46.5% | estimated ± 3.7 pp, low confidence |
| AA EnterpriseOps-Gym | 38.2% | estimated ± 7.2 pp, low confidence |
| AA Harvey LAB v1.0 | 88.4% | measured |
| AA Tau3 Banking | 47.1% | estimated ± 2.4 pp, low confidence |
| AA Terminal-Bench 2.1 | 80.0% | estimated ± 2.5 pp, medium confidence |
| ApprenticeBench | 0.0% | estimated ± 6.3 pp, low confidence |
| ARC-AGI-1 | 90.9% | estimated ± 6.5 pp, medium confidence |
| ARC-AGI-2 | 71.5% | estimated ± 13.8 pp, low confidence |
| BabaIsBench | 46.0% | measured |
| CursorBench 3.1 | 36.9% | estimated ± 3.4 pp, low confidence |
| CursorBench 3.2 | 56.8% | estimated ± 3.6 pp, high confidence |
| EBR-bench | 4.8% | estimated ± 8.1 pp, low confidence |
| Terminal-Bench 3.0 | 12.1% | estimated ± 3.2 pp, medium confidence |
| FrontierCode 1.1 Main | 28.4% | estimated ± 4.2 pp, medium confidence |
| FrontierSWE v2 | 13.9% | estimated ± 8.3 pp, medium confidence |
| Furniture Assembly | 32.9% | estimated ± 9.3 pp, medium confidence |
| MaxxBench | 49.1% | estimated ± 5.7 pp, medium confidence |
| MazeBench | 0.0% | measured |
| Mystery Game Puzzles | 8.0% | measured |
| NYT Connections | 65.1% | measured |
| OfficeQA Pro | 45.1% | measured |
| OSWorld-Verified | 70.1% | measured |
| ParseBench-1 | 62.7% | estimated ± 5.4 pp, low confidence |
| Terminal-Bench 4.1 | 7.6% | measured |
| Terminal-Bench Hard | 54.6% | estimated ± 2.8 pp, medium confidence |
| Vals SWE-bench | 75.0% | measured |
| Terminal-Bench 2.1 (Vals AI) | 53.6% | measured |
| VulcanBench v3 | 82.0% | estimated ± 5.6 pp, low confidence |
| CWE-Bench-AA | 41.7% | measured |
| AA Cyber Index | 22.4% | measured |
| DeepsecBench-AA | 9.8% | estimated ± 8.9 pp, low confidence |
| HealthBench Professional (length-adjusted, Artificial Analysis) | 51.6% | estimated ± 4.3 pp, low confidence |
| HealthBench Professional (raw, Artificial Analysis) | 35.8% | estimated ± 0.8 pp, low confidence |
| AA ITBench | 42.7% | estimated ± 4.0 pp, high confidence |
| Bug Hunt Bench | 11.6% | estimated ± 7.2 pp, low confidence |
| Chartography (tools) | 82.5% | estimated ± 4.1 pp, low confidence |
| CursorBench 4.0 | 34.8% | estimated ± 3.3 pp, medium confidence |
| CWE-bench v1 | 57.4% | estimated ± 5.1 pp, medium confidence |
| DeepSWE | 29.9% | estimated ± 6.4 pp, low confidence |
| DRACO | 86.9% | estimated ± 0.7 pp, low confidence |
| FrontierCode 2.0 | 20.9% | estimated ± 6.7 pp, medium confidence |
| FrontierPhysics | 1.7% | estimated ± 4.7 pp, high confidence |
| HLE w/o tools | 28.5% | estimated ± 4.9 pp, high confidence |
| OSWorld 2.0 | 4.6% | measured |
| OSWorld 2.1 | 59.0% | estimated ± 5.3 pp, low confidence |
| OSWorld 2.1 Offline | 20.7% | estimated ± 1.9 pp, low confidence |
| PostTrainBench v1.1 | 19.8% | estimated ± 3.7 pp, medium confidence |
| SpireClimb | 5.8% | estimated ± 7.2 pp, low confidence |
| StealthMark | 38.6% | measured |
| Terminal-Bench 4.0 | 15.8% | estimated ± 4.5 pp, medium confidence |
| Toolathlon-Verified | 52.5% | estimated ± 1.0 pp, low confidence |
| VulcanBench Frontier v4 | 77.3% | estimated ± 6.8 pp, low confidence |
| AA AutomationBench | 48.5% | estimated ± 8.2 pp, medium confidence |
| GDP.pdf | 17.0% | estimated ± 4.6 pp, high confidence |
| AA Terminal-Bench 4.0 | 7.2% | estimated ± 6.2 pp, medium confidence |
| ARC-AGI-3 | 0.0% | estimated ± 8.6 pp, low confidence |
| ARC-AGI-3 Provider Adapter | 2.7% | estimated ± 7.3 pp, medium confidence |
| cursorBench41 | 44.8% | estimated ± 3.7 pp, medium confidence |
| CWE-bench | 53.9% | estimated ± 7.9 pp, low confidence |
| MKA | 41.9% | estimated ± 5.7 pp, low confidence |
| SkillsBench official leaderboard (BenchFlow) | 53.0% | measured |
| BinaryAudit | 37.6% | estimated ± 12.0 pp, low confidence |
| CompileBench | 57.8% | estimated ± 10.0 pp, low confidence |
| OTelBench | 22.5% | estimated ± 6.6 pp, medium confidence |
| Vibe Code Bench | 52.2% | estimated ± 8.4 pp, medium confidence |
| FACTS Multimodal | 37.3% | estimated ± 4.0 pp, high confidence |
| Hermes Bench | 64.4% | estimated ± 2.9 pp, high confidence |
| Hermes Index | 37.2% | estimated ± 3.1 pp, high confidence |
| ParseBench-2 | 69.3% | estimated ± 5.0 pp, medium confidence |
| CharXiv | 78.8% | estimated ± 4.9 pp, high confidence |
| CharXiv w/o tools | 82.2% | estimated ± 1.2 pp, high confidence |
| ExploitBench | 16.5% | estimated ± 7.2 pp, low confidence |
| ExploitGym | 9.7% | estimated ± 2.7 pp, medium confidence |
| AI2D_TEST | 91.8% | estimated ± 1.3 pp, medium confidence |
| C-Eval | 91.2% | estimated ± 1.0 pp, medium confidence |
| Claw-Eval | 74.5% | measured |
| CountBench | 91.8% | estimated ± 2.0 pp, medium confidence |
| DeepPlanning | 47.3% | estimated ± 13.1 pp, low confidence |
| DynaMath | 84.7% | estimated ± 2.1 pp, medium confidence |
| ERQA | 57.1% | estimated ± 2.9 pp, high confidence |
| MathVision | 87.4% | estimated ± 3.0 pp, high confidence |
| MCP Atlas | 74.2% | measured |
| MLVU (M-Avg) | 86.0% | estimated ± 2.2 pp, high confidence |
| MMLU-Redux | 96.6% | estimated ± 0.8 pp, low confidence |
| MMMU-Pro | 78.1% | measured |
| NL2Repo | 42.1% | measured |
| OmniDocBench 1.5 | 91.6% | measured |
| PostTrainBench v1 | 18.2% | estimated ± 3.6 pp, high confidence |
| QwenClawBench | 61.9% | estimated ± 2.6 pp, high confidence |
| RealWorldQA | 83.7% | estimated ± 2.8 pp, high confidence |
| ScreenSpot Pro | 52.5% | estimated ± 8.4 pp, low confidence |
| SimpleVQA | 72.9% | estimated ± 7.1 pp, low confidence |
| SuperGPQA | 90.3% | estimated ± 9.3 pp, medium confidence |
| Terminal-Bench 2.0 | 61.5% | estimated ± 5.8 pp, medium confidence |
| Toolathlon | 53.2% | estimated ± 4.2 pp, high confidence |
| V* | 100.0% | estimated ± 7.5 pp, medium confidence |
| Video-MME (w/o subtitle) | 84.5% | estimated ± 1.3 pp, high confidence |
| Video-MME (with subtitle) | 85.4% | measured |
| VideoMMMU | 84.6% | measured |
| We-Math | 74.5% | estimated ± 3.5 pp, medium confidence |
| AA LiveCodeBench | 88.9% | estimated ± 2.6 pp, high confidence |
| HealthBench Hard | 21.0% | estimated ± 8.6 pp, low confidence |
| LiveCodeBench Pro | 83.0% | estimated ± 6.0 pp, low confidence |
| MedXpertQA (MM) | 70.4% | estimated ± 2.4 pp, high confidence |
| MMLU-Pro (Arcee) | 91.0% | estimated ± 4.0 pp, low confidence |
| React Native Evals | 79.9% | estimated ± 2.7 pp, medium confidence |
| ResearchClawBench | 19.8% | measured |
| SWE-Rebench | 61.8% | estimated ± 3.6 pp, medium confidence |
| PostTrainBench | 17.2% | estimated ± 3.7 pp, medium confidence |
| DAYJOB | 0.0% | estimated ± 3.2 pp, high confidence |
| DAYJOB Healthcare | 0.7% | estimated ± 4.2 pp, high confidence |
| Finance Agent v2 | 56.1% | estimated ± 3.8 pp, medium confidence |
| BioMysteryBench (human-solvable) | 81.7% | estimated ± 0.6 pp, low confidence |
| FrontierCode 1.1 Extended | 55.7% | estimated ± 3.0 pp, low confidence |
| HealthBench (raw) | 47.8% | estimated ± 6.0 pp, medium confidence |
| HealthBench (length-adjusted) | 53.2% | estimated ± 3.4 pp, medium confidence |
| HealthBench Professional | 56.7% | estimated ± 4.2 pp, low confidence |
| HealthBench Professional (raw) | 58.4% | estimated ± 4.5 pp, high confidence |
| HLE-Verified | 53.0% | estimated ± 2.0 pp, medium confidence |
| LABBench2 | 79.9% | estimated ± 1.8 pp, low confidence |
| ScarfBench | 25.8% | estimated ± 9.1 pp, low confidence |
| BlenderWorld (text-only) | 70.0% | estimated ± 2.8 pp, medium confidence |
| Cookie-Bench | 76.9% | estimated ± 1.3 pp, medium confidence |
| STEMBench | 2.5% | estimated ± 0.2 pp, medium confidence |
| SimpleQA | 52.3% | estimated ± 4.3 pp, low confidence |
| LiveCodeBench | 79.6% | estimated ± 14.6 pp, low confidence |
| Agents' Last Exam | 35.4% | estimated ± 11.0 pp, low confidence |
| BabyVision w/ Python | 88.8% | estimated ± 3.8 pp, low confidence |
| OpenHarmony Bench v1.0 | 48.4% | measured |
| SEC-Bench Pro | 44.9% | estimated ± 9.3 pp, medium confidence |
| BabyVision | 61.0% | estimated ± 9.1 pp, low confidence |
| MMVU | 77.2% | estimated ± 2.7 pp, medium confidence |
| PerceptionBench | 59.7% | estimated ± 4.9 pp, low confidence |
| MRCRv2 | 85.9% | estimated ± 11.3 pp, low confidence |
| CC-OCR | 82.5% | estimated ± 3.7 pp, medium confidence |
| MMLongBench-Doc | 61.5% | estimated ± 2.6 pp, low confidence |
| OCRBench V2 | 70.7% | estimated ± 2.1 pp, medium confidence |
| BioBasicsBench | 33.9% | estimated ± 2.8 pp, medium confidence |
| LifeVal | 74.9% | estimated ± 1.6 pp, medium confidence |
| MMMU-Pro w/ Python | 81.5% | estimated ± 2.0 pp, high confidence |
| ExploitBench Internal Port | 0.0% | estimated ± 1.4 pp, low confidence |
| PinchBench | 90.4% | estimated ± 3.6 pp, low confidence |