| Benchmark | Gemini 3 Pro | o1 |
|---|---|---|
| AA Agentic Index | 52 | 31.1 |
| AA Intelligence | 40.6 | 24 |
| AA-LCR | 73 | 63.3 |
| AA-Omniscience | 15.3 | -10.6 |
| AIME 2025 | 100 | 81.7 |
| arena_vision | 1290 | 1193 |
| Artificial Analysis Coding Index | 46.5 | 39.7 |
| critpt | 9.1 | 0.3 |
| Fortress | 41.7 | 19.4 |
| gdpval | 34.2 | 11.5 |
| GPQA Diamond | 91.9 | 75.7 |
| HLE | 45.8 | 7.7 |
| IFBench | 70.4 | 70.3 |
| livebench | 73.4 | 52.3 |
| MathVista | 89.8 | 71.8 |
| metr_hcast | 71.3 | 42.6 |
| metr_re_bench | 35 | 0 |
| metr_swaa | 99.5 | 100 |
| mmmlu | 91.8 | 87.7 |
| OmniScience Accuracy | 55.8 | 34.7 |
| OmniScience Non-Hallucination | 10 | 30.7 |
| scicode | 56.1 | 35.8 |
| SEAL VISTA | 51.5 | 45.3 |
| simplebench | 76.4 | 41.7 |
| simpleqa | 72.1 | 47 |
| SWE-bench Verified | 78 | 48.9 |
| Terminal-Bench Hard | 56.9 | 12.9 |
| τ²-Bench Telecom (AA run) | 98 | 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.