| Benchmark | Claude 3.7 Sonnet | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 37 | 57.8 |
| AA Intelligence | 27.6 | 61 |
| AA-LCR | 62.3 | 77.7 |
| AA-Omniscience | -0.7 | 22 |
| arc_agi_1 | 13.6 | 97.5 |
| ARC-AGI-2 | 0 | 92.5 |
| Artificial Analysis Coding Index | 36.4 | 78.3 |
| critpt | 0.9 | 32.3 |
| gdpval | 27.4 | 61.1 |
| GPQA Diamond | 84.8 | 94.6 |
| HLE | 9.7 | 49.5 |
| IFBench | 48.3 | 72.7 |
| MMMU-Pro | 60.1 | 84.6 |
| OmniScience Accuracy | 28.2 | 59.4 |
| OmniScience Non-Hallucination | 60 | 10.6 |
| scicode | 40.3 | 56.9 |
| simplebench | 46.4 | 64.8 |
| Terminal-Bench Hard | 21.2 | 65.9 |
| τ²-Bench Telecom (AA run) | 54.7 | 85.1 |
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.