| Benchmark | Claude 3.7 Sonnet | Claude Opus 4.6 |
|---|---|---|
| AA Agentic Index | 37 | 67.6 |
| AA Intelligence | 27.6 | 44.9 |
| AA-LCR | 62.3 | 74.3 |
| AA-Omniscience | -0.7 | 13.7 |
| AIME 2025 | 54.8 | 99.8 |
| arc_agi_1 | 13.6 | 93 |
| ARC-AGI-2 | 0 | 68.8 |
| arena_vision | 1195 | 1300 |
| Artificial Analysis Coding Index | 36.4 | 48.1 |
| critpt | 0.9 | 12.6 |
| Fortress | 38 | 20.5 |
| gdpval | 27.4 | 55.9 |
| GPQA Diamond | 84.8 | 91.3 |
| HLE | 9.7 | 53.1 |
| IFBench | 48.3 | 62.5 |
| metr_hcast | 54.2 | 77 |
| metr_re_bench | 0 | 97.4 |
| metr_swaa | 98.7 | 98.5 |
| mmmlu | 86.1 | 91.1 |
| MMMU-Pro | 60.1 | 77.3 |
| multichallenge | 51.6 | 56 |
| MultiNRC | 27.8 | 57.1 |
| OmniScience Accuracy | 28.2 | 47 |
| OmniScience Non-Hallucination | 60 | 37.2 |
| scicode | 40.3 | 52 |
| SEAL VISTA | 48.2 | 46.1 |
| simplebench | 46.4 | 67.6 |
| swe_bench_bash | 52.8 | 75.6 |
| SWE-bench Verified | 70.3 | 80.8 |
| Terminal-Bench Hard | 21.2 | 65.4 |
| τ²-Bench Telecom (AA run) | 54.7 | 99.3 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.