Anthropic · model
Claude Sonnet 4.5 Thinking benchmark scores
As of 2026-10-07, Claude Sonnet 4.5 Thinking (Anthropic) has measured scores on 3 benchmarks and estimated scores on 24 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 85.5% | estimated ± 2.7 pp, high confidence |
| AA-HLE | 29.7% | estimated ± 6.2 pp, medium confidence |
| AA-SciCode | 49.0% | estimated ± 3.8 pp, high confidence |
| Artificial Analysis Intelligence Index | 21.7% | estimated ± 7.0 pp, medium confidence |
| CritPt | 9.8% | estimated ± 5.7 pp, medium confidence |
| GPQA Diamond | 84.3% | estimated ± 1.4 pp, medium confidence |
| HLE | 43.8% | estimated ± 6.3 pp, medium confidence |
| AA-Omniscience Accuracy | 33.0% | estimated ± 6.2 pp, medium confidence |
| AA Coding Index | 51.0% | estimated ± 6.8 pp, medium confidence |
| SWE-bench Verified | 75.9% | estimated ± 4.4 pp, high confidence |
| GPQA | 84.8% | estimated ± 1.6 pp, high confidence |
| ARC-AGI-1 | 63.7% | measured |
| ARC-AGI-2 | 13.6% | measured |
| FrontierCode 1.1 Main | 9.4% | estimated ± 4.3 pp, low confidence |
| Vals GPQA Diamond | 80.7% | estimated ± 2.2 pp, high confidence |
| Vals MMLU-Pro | 82.3% | estimated ± 1.8 pp, high confidence |
| Vals SWE-bench | 73.3% | estimated ± 6.3 pp, medium confidence |
| HLE w/o tools | 28.2% | estimated ± 6.0 pp, medium confidence |
| PostTrainBench v1.1 | 19.9% | estimated ± 10.5 pp, low confidence |
| Vibe Code Bench | 22.6% | measured |
| React Native Evals | 70.7% | estimated ± 3.0 pp, low confidence |
| SWE-Rebench | 42.8% | estimated ± 6.5 pp, low confidence |
| ARC-AGI-3 | 0.0% | estimated ± 13.8 pp, low confidence |
| HealthBench (length-adjusted) | 43.6% | estimated ± 2.5 pp, low confidence |
| HealthBench Professional | 55.1% | estimated ± 4.6 pp, medium confidence |
| HLE-Verified | 35.3% | estimated ± 0.7 pp, low confidence |
| LABBench2 | 22.7% | estimated ± 1.3 pp, low confidence |