| Benchmark | Claude Sonnet 4.6 | o1 |
|---|---|---|
| AA Agentic Index | 61.6 | 31.1 |
| AA Intelligence | 48.4 | 24 |
| AA-LCR | 74 | 63.3 |
| AA-Omniscience | 12.2 | -10.6 |
| arena_vision | 1278 | 1193 |
| Artificial Analysis Coding Index | 63 | 39.7 |
| critpt | 3.1 | 0.3 |
| gdpval | 54.8 | 11.5 |
| GPQA Diamond | 89.9 | 75.7 |
| harmbench | 24.2 | 96.3 |
| HLE | 49 | 7.7 |
| IFBench | 56.6 | 70.3 |
| livebench | 75.5 | 52.3 |
| mmmlu | 89.3 | 87.7 |
| OmniScience Accuracy | 40.9 | 34.7 |
| OmniScience Non-Hallucination | 51.6 | 30.7 |
| scicode | 47 | 35.8 |
| SWE-bench Verified | 79.6 | 48.9 |
| Terminal-Bench Hard | 59.1 | 12.9 |
| τ²-Bench Telecom (AA run) | 97.9 | 62.6 |
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.