Anthropic · model
Claude Mythos 5 benchmark scores
As of 2026-11-21, Claude Mythos 5 (Anthropic) has measured scores on 14 benchmarks and estimated scores on 126 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 93.6% | estimated ± 2.7 pp, high confidence |
| AA-HLE | 57.9% | estimated ± 4.6 pp, high confidence |
| AA-SciCode | 60.8% | estimated ± 3.6 pp, high confidence |
| Artificial Analysis Intelligence Index | 51.0% | estimated ± 5.6 pp, medium confidence |
| CritPt | 24.5% | estimated ± 4.0 pp, high confidence |
| GDPval-AA | 50.8% | estimated ± 5.0 pp, high confidence |
| GPQA Diamond | 94.1% | estimated ± 1.2 pp, high confidence |
| HLE | 64.5% | measured |
| LiveCodeBench v6 | 96.9% | estimated ± 2.7 pp, medium confidence |
| AA-Omniscience Accuracy | 60.9% | estimated ± 7.7 pp, medium confidence |
| SciCode | 54.3% | estimated ± 3.6 pp, medium confidence |
| Terminal-Bench 2.1 | 88.0% | measured |
| BrowseComp | 88.0% | measured |
| HLE w/ tools | 61.5% | estimated ± 4.9 pp, medium confidence |
| VITA-Bench | 46.8% | estimated ± 6.7 pp, low confidence |
| τ²-bench results | 93.5% | estimated ± 11.8 pp, low confidence |
| AA Coding Index | 78.2% | estimated ± 5.3 pp, medium confidence |
| AA-MMMU-Pro | 84.7% | estimated ± 2.9 pp, high confidence |
| APEX-Agents-AA | 37.7% | estimated ± 4.4 pp, high confidence |
| DeepSearchQA | 90.7% | estimated ± 6.8 pp, low confidence |
| SWE-bench Verified | 95.5% | measured |
| AutomationBench | 45.1% | estimated ± 10.4 pp, low confidence |
| CyberGym | 83.8% | measured |
| JobBench | 52.4% | estimated ± 8.6 pp, low confidence |
| SkillsBench | 60.6% | estimated ± 4.1 pp, medium confidence |
| SWE-bench Pro | 80.3% | measured |
| WideResearch | 83.0% | estimated ± 1.6 pp, medium confidence |
| AA Agentic Index | 48.5% | estimated ± 5.8 pp, medium confidence |
| MMLU-Pro | 86.0% | estimated ± 2.5 pp, low confidence |
| GPQA | 94.1% | measured |
| MMLU | 88.1% | estimated ± 4.2 pp, medium confidence |
| Vals GPQA Diamond | 94.4% | estimated ± 2.9 pp, high confidence |
| Vals LiveCodeBench | 88.3% | estimated ± 4.7 pp, medium confidence |
| Vals MMLU-Pro | 91.3% | estimated ± 1.8 pp, high confidence |
| MMMLU | 91.5% | estimated ± 1.7 pp, medium confidence |
| MMMU | 93.1% | estimated ± 3.7 pp, low confidence |
| Chess Puzzles | 54.4% | estimated ± 8.0 pp, medium confidence |
| AA MMLU-Pro | 87.3% | estimated ± 1.2 pp, low confidence |
| Gert Labs | 69.4% | estimated ± 5.5 pp, low confidence |
| AA Harvey LAB v1.0 | 92.4% | estimated ± 9.8 pp, low confidence |
| AA Tau3 Banking | 41.5% | estimated ± 3.3 pp, medium confidence |
| AA Terminal-Bench 2.1 | 86.2% | estimated ± 2.6 pp, high confidence |
| ApprenticeBench | 24.1% | estimated ± 4.0 pp, high confidence |
| ARC-AGI-1 | 96.9% | estimated ± 3.9 pp, high confidence |
| ARC-AGI-2 | 79.2% | estimated ± 10.9 pp, low confidence |
| CursorBench 3.1 | 66.9% | estimated ± 5.6 pp, low confidence |
| CursorBench 3.2 | 71.4% | estimated ± 3.6 pp, high confidence |
| Terminal-Bench 3.0 | 23.0% | estimated ± 9.1 pp, low confidence |
| FrontierCode 1.1 Main | 54.9% | estimated ± 4.2 pp, medium confidence |
| FrontierSWE v2 | 50.7% | estimated ± 8.3 pp, medium confidence |
| Furniture Assembly | 66.3% | estimated ± 6.5 pp, medium confidence |
| MazeBench | 5.3% | estimated ± 6.4 pp, low confidence |
| MirrorCode | 66.3% | estimated ± 14.9 pp, low confidence |
| Mystery Game Puzzles | 54.8% | estimated ± 7.5 pp, medium confidence |
| NYT Connections | 94.2% | estimated ± 10.3 pp, low confidence |
| OfficeQA Pro | 67.0% | estimated ± 2.5 pp, medium confidence |
| OSWorld-Verified | 85.0% | measured |
| ParseBench-1 | 71.2% | estimated ± 1.0 pp, low confidence |
| Terminal-Bench 4.1 | 24.9% | estimated ± 10.9 pp, low confidence |
| Vals SWE-bench | 97.0% | estimated ± 6.6 pp, medium confidence |
| Terminal-Bench 2.1 (Vals AI) | 73.7% | estimated ± 7.3 pp, medium confidence |
| CWE-Bench-AA | 54.3% | estimated ± 6.4 pp, medium confidence |
| AA Cyber Index | 41.6% | estimated ± 13.5 pp, low confidence |
| DeepsecBench-AA | 21.9% | estimated ± 9.0 pp, low confidence |
| HealthBench Professional (length-adjusted, Artificial Analysis) | 60.2% | estimated ± 4.4 pp, medium confidence |
| HealthBench Professional (raw, Artificial Analysis) | 69.4% | estimated ± 2.6 pp, medium confidence |
| AA ITBench | 49.2% | estimated ± 4.1 pp, high confidence |
| Bug Hunt Bench | 39.7% | estimated ± 12.1 pp, low confidence |
| CursorBench 4.0 | 50.9% | estimated ± 3.3 pp, high confidence |
| DeepSWE | 64.6% | estimated ± 11.2 pp, low confidence |
| DRACO | 84.1% | estimated ± 2.9 pp, medium confidence |
| FrontierCode 2.0 | 34.4% | estimated ± 6.7 pp, medium confidence |
| FrontierPhysics | 7.0% | estimated ± 2.7 pp, high confidence |
| HLE w/o tools | 59.0% | measured |
| OfficeQA | 78.4% | estimated ± 2.7 pp, medium confidence |
| OSWorld 2.0 | 19.9% | estimated ± 3.6 pp, high confidence |
| PostTrainBench v1.1 | 40.2% | estimated ± 4.4 pp, high confidence |
| SpireClimb | 3.5% | estimated ± 8.2 pp, low confidence |
| StealthMark | 56.5% | estimated ± 6.1 pp, medium confidence |
| SWE Multilingual | 92.2% | measured |
| SWE-bench Multimodal | 54.9% | measured |
| Toolathlon-Verified | 73.5% | estimated ± 2.2 pp, low confidence |
| VulcanBench Frontier v4 | 89.2% | estimated ± 8.3 pp, low confidence |
| AA Terminal-Bench 4.0 | 38.1% | estimated ± 11.0 pp, low confidence |
| ARC-AGI-3 | 26.5% | estimated ± 7.2 pp, medium confidence |
| Vibe Code Bench | 66.3% | estimated ± 8.4 pp, low confidence |
| FACTS Multimodal | 47.2% | estimated ± 2.7 pp, medium confidence |
| CharXiv | 93.5% | measured |
| CharXiv w/o tools | 88.9% | measured |
| ExploitBench | 78.0% | measured |
| ExploitGym | 22.5% | estimated ± 9.2 pp, low confidence |
| AI2D_TEST | 93.5% | estimated ± 0.9 pp, low confidence |
| C-Eval | 93.5% | estimated ± 1.2 pp, medium confidence |
| Claw-Eval | 71.9% | estimated ± 6.2 pp, low confidence |
| CountBench | 99.4% | estimated ± 3.7 pp, low confidence |
| DynaMath | 88.5% | estimated ± 0.8 pp, low confidence |
| ERQA | 78.9% | estimated ± 2.9 pp, low confidence |
| INCLUDE | 89.8% | estimated ± 2.0 pp, low confidence |
| MathVision | 93.1% | estimated ± 3.3 pp, high confidence |
| MCP Atlas | 79.4% | estimated ± 10.3 pp, low confidence |
| MLVU (M-Avg) | 90.1% | estimated ± 1.5 pp, high confidence |
| MMLU-Redux | 95.4% | estimated ± 1.7 pp, medium confidence |
| MMMU-Pro | 82.3% | estimated ± 3.1 pp, high confidence |
| NL2Repo | 71.7% | estimated ± 6.1 pp, low confidence |
| PostTrainBench v1 | 46.4% | estimated ± 4.6 pp, medium confidence |
| RealWorldQA | 88.5% | estimated ± 1.5 pp, high confidence |
| ScreenSpot Pro | 88.5% | estimated ± 6.8 pp, medium confidence |
| SuperGPQA | 100.0% | estimated ± 7.2 pp, low confidence |
| Toolathlon | 58.3% | estimated ± 4.2 pp, high confidence |
| Video-MME (with subtitle) | 92.1% | estimated ± 2.9 pp, high confidence |
| VideoMMMU | 88.2% | estimated ± 1.5 pp, medium confidence |
| We-Math | 84.6% | estimated ± 7.9 pp, low confidence |
| HealthBench Hard | 36.4% | estimated ± 5.1 pp, low confidence |
| LiveCodeBench Pro | 97.4% | estimated ± 6.0 pp, low confidence |
| MedXpertQA (MM) | 80.4% | estimated ± 6.4 pp, low confidence |
| MedXpertQA (Text) | 74.6% | estimated ± 8.9 pp, low confidence |
| React Native Evals | 86.6% | estimated ± 2.7 pp, low confidence |
| ResearchClawBench | 20.3% | estimated ± 2.1 pp, low confidence |
| SWE-Rebench | 61.8% | estimated ± 3.6 pp, low confidence |
| PostTrainBench | 37.1% | estimated ± 3.7 pp, high confidence |
| DAYJOB | 5.3% | estimated ± 1.1 pp, medium confidence |
| DAYJOB Healthcare | 1.5% | estimated ± 1.5 pp, medium confidence |
| FrontierCode 1.1 Extended | 61.5% | estimated ± 3.0 pp, medium confidence |
| HealthBench Professional | 64.5% | estimated ± 4.2 pp, medium confidence |
| SimpleQA | 38.6% | estimated ± 13.4 pp, low confidence |
| LiveCodeBench | 87.9% | estimated ± 14.6 pp, low confidence |
| Agents' Last Exam | 28.1% | estimated ± 3.3 pp, medium confidence |
| BabyVision w/ Python | 92.1% | estimated ± 3.7 pp, low confidence |
| OpenHarmony Bench v1.0 | 60.9% | estimated ± 4.9 pp, low confidence |
| SEC-Bench Pro | 78.4% | estimated ± 3.9 pp, high confidence |
| BabyVision | 85.1% | estimated ± 7.4 pp, low confidence |
| MMVU | 82.1% | estimated ± 3.6 pp, medium confidence |
| PerceptionBench | 66.2% | estimated ± 3.8 pp, medium confidence |
| ZeroBench | 29.8% | estimated ± 7.1 pp, low confidence |
| LVBench | 86.6% | estimated ± 8.2 pp, low confidence |
| BioBasicsBench | 41.0% | estimated ± 2.2 pp, medium confidence |
| LifeVal | 78.3% | estimated ± 1.6 pp, medium confidence |
| ExploitBench Internal Port | 11.9% | estimated ± 12.3 pp, low confidence |
| MathVision w/ Python | 97.7% | estimated ± 1.0 pp, low confidence |
| PinchBench | 90.6% | estimated ± 3.6 pp, low confidence |