Anthropic · model
Claude 4 Sonnet benchmark scores
As of 2026-11-22, Claude 4 Sonnet (Anthropic) has measured scores on 13 benchmarks and estimated scores on 103 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 68.3% | measured |
| AA-HLE | 4.3% | measured |
| AA-SciCode | 45.0% | estimated ± 3.8 pp, high confidence |
| AIME26 | 37.6% | estimated ± 4.3 pp, low confidence |
| Artificial Analysis Intelligence Index | 16.6% | measured |
| CritPt | 1.1% | measured |
| GDPval-AA | 26.1% | estimated ± 6.2 pp, medium confidence |
| GPQA Diamond | 68.8% | estimated ± 2.4 pp, medium confidence |
| HLE | 16.7% | estimated ± 8.5 pp, low confidence |
| IFBench | 54.9% | estimated ± 6.6 pp, medium confidence |
| AA-LCR | 44.0% | measured |
| AA-Omniscience Accuracy | 22.7% | measured |
| SciCode | 44.2% | estimated ± 3.6 pp, high confidence |
| BrowseComp | 67.7% | estimated ± 6.2 pp, medium confidence |
| HLE w/ tools | 47.0% | estimated ± 5.7 pp, medium confidence |
| IFEval | 90.3% | estimated ± 2.8 pp, high confidence |
| LongBench v2 | 51.7% | estimated ± 5.5 pp, medium confidence |
| τ²-bench results | 52.3% | measured |
| AA Coding Index | 45.5% | estimated ± 5.3 pp, medium confidence |
| AA-MMMU-Pro | 62.4% | measured |
| APEX-Agents-AA | 21.9% | estimated ± 3.6 pp, medium confidence |
| BioMysteryBench (human-difficult) | 18.0% | estimated ± 6.8 pp, low confidence |
| DeepSearchQA | 76.0% | estimated ± 8.2 pp, low confidence |
| SWE-bench Verified | 72.7% | measured |
| BFCL v4 | 44.9% | estimated ± 8.5 pp, low confidence |
| CyberGym | 51.4% | estimated ± 11.0 pp, low confidence |
| JobBench | 18.4% | measured |
| SWE-bench Pro | 50.4% | estimated ± 4.5 pp, high confidence |
| WideResearch | 74.4% | estimated ± 1.2 pp, medium confidence |
| AA Agentic Index | 16.7% | estimated ± 9.4 pp, medium confidence |
| MMLU-Pro | 84.6% | estimated ± 5.0 pp, medium confidence |
| AA-IFBench | 45.4% | measured |
| AIME 2024 | 48.6% | estimated ± 2.6 pp, high confidence |
| GPQA | 70.8% | estimated ± 6.1 pp, medium confidence |
| MMLU | 90.3% | estimated ± 0.9 pp, high confidence |
| Vals GPQA Diamond | 67.4% | estimated ± 3.2 pp, high confidence |
| Vals LiveCodeBench | 73.8% | estimated ± 10.0 pp, low confidence |
| Vals MMLU-Pro | 81.0% | estimated ± 2.9 pp, high confidence |
| MATH-500 | 91.4% | estimated ± 2.0 pp, high confidence |
| MMMLU | 86.5% | estimated ± 0.7 pp, medium confidence |
| MMMU | 74.8% | estimated ± 2.7 pp, high confidence |
| AIME 2025 | 33.1% | measured |
| AA MMLU-Pro | 84.5% | estimated ± 0.8 pp, medium confidence |
| Gert Labs | 27.7% | measured |
| AA EnterpriseOps-Gym | 41.3% | estimated ± 1.2 pp, low confidence |
| ApprenticeBench | 0.2% | estimated ± 6.0 pp, low confidence |
| ARC-AGI-1 | 75.6% | estimated ± 7.8 pp, low confidence |
| ARC-AGI-2 | 32.7% | estimated ± 14.4 pp, low confidence |
| FrontierCode 1.1 Main | 14.6% | estimated ± 4.2 pp, low confidence |
| MazeBench | 0.0% | estimated ± 2.7 pp, high confidence |
| OSWorld-Verified | 61.0% | estimated ± 8.4 pp, low confidence |
| ParseBench-1 | 47.7% | estimated ± 6.2 pp, medium confidence |
| Vals SWE-bench | 66.3% | estimated ± 6.6 pp, medium confidence |
| CWE-Bench-AA | 38.3% | estimated ± 8.1 pp, medium confidence |
| AA Cyber Index | 15.6% | estimated ± 12.4 pp, low confidence |
| DeepsecBench-AA | 8.4% | estimated ± 8.2 pp, low confidence |
| HealthBench Professional (length-adjusted, Artificial Analysis) | 50.9% | estimated ± 4.3 pp, low confidence |
| HealthBench Professional (raw, Artificial Analysis) | 51.9% | estimated ± 1.1 pp, low confidence |
| AA ITBench | 36.4% | estimated ± 4.6 pp, medium confidence |
| CWE-bench v1 | 37.9% | estimated ± 5.7 pp, medium confidence |
| DeepSWE | 34.1% | estimated ± 12.7 pp, low confidence |
| HLE w/o tools | 29.1% | estimated ± 4.9 pp, high confidence |
| OSWorld 2.1 | 7.7% | estimated ± 3.6 pp, medium confidence |
| Toolathlon-Verified | 61.7% | estimated ± 3.0 pp, high confidence |
| AA AutomationBench | 42.0% | estimated ± 9.9 pp, low confidence |
| GDP.pdf | 16.4% | estimated ± 4.1 pp, low confidence |
| ARC-AGI-3 | 1.3% | estimated ± 9.8 pp, low confidence |
| ARC-AGI-3 Provider Adapter | 2.7% | estimated ± 7.3 pp, low confidence |
| CWE-bench | 31.6% | estimated ± 6.2 pp, low confidence |
| MKA | 32.8% | estimated ± 4.6 pp, medium confidence |
| Vibe Code Bench | 0.4% | estimated ± 8.4 pp, medium confidence |
| Hermes Index | 19.1% | estimated ± 5.7 pp, low confidence |
| ParseBench-2 | 48.8% | estimated ± 6.2 pp, medium confidence |
| CharXiv | 72.2% | estimated ± 6.3 pp, low confidence |
| CharXiv w/o tools | 71.5% | estimated ± 2.5 pp, medium confidence |
| ExploitGym | 0.0% | estimated ± 8.3 pp, low confidence |
| AI2D_TEST | 88.0% | estimated ± 2.0 pp, medium confidence |
| C-Eval | 90.5% | estimated ± 1.0 pp, medium confidence |
| Claw-Eval | 55.3% | estimated ± 4.5 pp, medium confidence |
| CountBench | 85.8% | estimated ± 2.9 pp, low confidence |
| DynaMath | 0.4% | estimated ± 3.0 pp, low confidence |
| ERQA | 41.2% | estimated ± 3.1 pp, medium confidence |
| HMMT Feb 2026 | 25.6% | estimated ± 11.0 pp, low confidence |
| MathVision | 62.2% | estimated ± 3.1 pp, medium confidence |
| MCP Atlas | 42.2% | estimated ± 8.6 pp, medium confidence |
| MMLU-Redux | 93.8% | estimated ± 0.8 pp, medium confidence |
| MMMU-Pro | 63.7% | estimated ± 1.5 pp, high confidence |
| NL2Repo | 31.0% | estimated ± 6.2 pp, medium confidence |
| PostTrainBench v1 | 12.0% | estimated ± 5.5 pp, low confidence |
| RealWorldQA | 74.2% | estimated ± 2.9 pp, low confidence |
| ScreenSpot Pro | 66.1% | estimated ± 10.8 pp, low confidence |
| SuperGPQA | 65.1% | estimated ± 9.3 pp, medium confidence |
| Toolathlon | 30.9% | estimated ± 7.4 pp, low confidence |
| V* | 0.0% | estimated ± 7.5 pp, low confidence |
| Video-MME (w/o subtitle) | 75.6% | estimated ± 1.7 pp, medium confidence |
| Video-MME (with subtitle) | 76.2% | estimated ± 3.7 pp, low confidence |
| HealthBench Hard | 10.4% | estimated ± 8.6 pp, low confidence |
| MedXpertQA (MM) | 7.6% | estimated ± 6.5 pp, low confidence |
| MMLU-Pro (Arcee) | 81.3% | estimated ± 4.0 pp, medium confidence |
| React Native Evals | 76.4% | estimated ± 2.7 pp, medium confidence |
| ResearchClawBench | 15.2% | estimated ± 2.6 pp, low confidence |
| SWE-Rebench | 57.0% | estimated ± 3.6 pp, medium confidence |
| Finance Agent v2 | 55.1% | estimated ± 2.4 pp, low confidence |
| BioMysteryBench (human-solvable) | 81.7% | estimated ± 0.6 pp, low confidence |
| HealthBench (length-adjusted) | 54.1% | estimated ± 3.7 pp, medium confidence |
| HealthBench Professional | 56.7% | estimated ± 4.2 pp, low confidence |
| HLE-Verified | 17.8% | estimated ± 7.1 pp, low confidence |
| STEMBench | 2.4% | estimated ± 0.2 pp, low confidence |
| SimpleQA | 21.8% | estimated ± 4.3 pp, medium confidence |
| LiveCodeBench | 74.7% | estimated ± 14.6 pp, low confidence |
| Agents' Last Exam | 29.6% | estimated ± 9.7 pp, low confidence |
| MRCRv2 | 22.8% | estimated ± 11.3 pp, low confidence |
| MMLongBench-Doc | 58.7% | estimated ± 2.6 pp, low confidence |
| BioBasicsBench | 34.1% | estimated ± 3.8 pp, low confidence |
| LifeVal | 4.5% | estimated ± 5.0 pp, low confidence |
| MMMU-Pro w/ Python | 66.9% | estimated ± 2.0 pp, medium confidence |