| Benchmark | Claude 3.7 Sonnet | Claude Opus 4.7 |
|---|---|---|
| AA Agentic Index | 37 | 64.6 |
| AA Intelligence | 27.6 | 55 |
| AA-LCR | 62.3 | 75.3 |
| AA-Omniscience | -0.7 | 27.3 |
| arc_agi_1 | 13.6 | 92 |
| ARC-AGI-2 | 0 | 75.8 |
| arena_vision | 1195 | 1304 |
| Artificial Analysis Coding Index | 36.4 | 73.6 |
| critpt | 0.9 | 12 |
| gdpval | 27.4 | 58.6 |
| GPQA Diamond | 84.8 | 94.2 |
| harmbench | 84.3 | 17.5 |
| HLE | 9.7 | 54.7 |
| IFBench | 48.3 | 58.6 |
| mmmlu | 86.1 | 91.5 |
| MMMU-Pro | 60.1 | 78.8 |
| OmniScience Accuracy | 28.2 | 48.9 |
| OmniScience Non-Hallucination | 60 | 57.7 |
| scicode | 40.3 | 54.5 |
| simplebench | 46.4 | 61.7 |
| SWE-bench Verified | 70.3 | 87.6 |
| Terminal-Bench Hard | 21.2 | 54.5 |
| τ²-Bench Telecom (AA run) | 54.7 | 88.6 |
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.