| Benchmark | Claude 3.7 Sonnet | Gemini 3.5 Flash |
|---|---|---|
| AA Agentic Index | 37 | 70.4 |
| AA Intelligence | 27.6 | 52 |
| AA-LCR | 62.3 | 81 |
| AA-Omniscience | -0.7 | 21.2 |
| arc_agi_1 | 13.6 | 92.5 |
| ARC-AGI-2 | 0 | 72.1 |
| Artificial Analysis Coding Index | 36.4 | 70.1 |
| critpt | 0.9 | 13.1 |
| gdpval | 27.4 | 57.8 |
| GPQA Diamond | 84.8 | 92.2 |
| HLE | 9.7 | 42.7 |
| IFBench | 48.3 | 76.3 |
| MMMU-Pro | 60.1 | 84.3 |
| OmniScience Accuracy | 28.2 | 51.4 |
| OmniScience Non-Hallucination | 60 | 38.2 |
| scicode | 40.3 | 53.1 |
| simplebench | 46.4 | 76.7 |
| Terminal-Bench Hard | 21.2 | 46.2 |
| τ²-Bench Telecom (AA run) | 54.7 | 95.6 |
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.