Anthropic · model
Claude Haiku 4.5 Thinking benchmark scores
As of 2026-11-22, Claude Haiku 4.5 Thinking (Anthropic) has measured scores on 3 benchmarks and estimated scores on 21 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-SciCode | 47.1% | estimated ± 3.8 pp, high confidence |
| AA Coding Index | 45.4% | estimated ± 6.9 pp, medium confidence |
| AA-MMMU-Pro | 65.0% | estimated ± 6.1 pp, medium confidence |
| SWE-bench Verified | 73.3% | estimated ± 4.5 pp, high confidence |
| SWE-bench Pro | 53.9% | estimated ± 2.9 pp, medium confidence |
| Mystery Game Puzzles | 14.5% | estimated ± 4.7 pp, medium confidence |
| OfficeQA Pro | 48.5% | estimated ± 7.0 pp, low confidence |
| ParseBench-1 | 51.2% | measured |
| Vals SWE-bench | 69.1% | estimated ± 7.1 pp, medium confidence |
| PostTrainBench v1.1 | 19.9% | estimated ± 10.5 pp, low confidence |
| SWE-bench Multimodal | 0.1% | estimated ± 5.6 pp, low confidence |
| CompileBench | 73.3% | measured |
| Vibe Code Bench | 11.4% | measured |
| ParseBench-2 | 61.3% | estimated ± 4.3 pp, high confidence |
| CharXiv | 80.5% | estimated ± 5.6 pp, low confidence |
| CharXiv w/o tools | 71.1% | estimated ± 1.8 pp, low confidence |
| MMMU-Pro | 71.9% | estimated ± 6.9 pp, medium confidence |
| PostTrainBench v1 | 19.0% | estimated ± 3.1 pp, low confidence |
| ScreenSpot Pro | 74.3% | estimated ± 10.1 pp, low confidence |
| AA LiveCodeBench | 85.1% | estimated ± 3.4 pp, low confidence |
| React Native Evals | 57.0% | estimated ± 3.0 pp, low confidence |
| SWE-Rebench | 2.2% | estimated ± 6.5 pp, low confidence |
| PostTrainBench | 17.0% | estimated ± 8.3 pp, low confidence |
| MMMU-Pro w/ Python | 75.2% | estimated ± 2.7 pp, medium confidence |