| Benchmark | Gemini 2.5 Pro | o1 |
|---|---|---|
| AA Agentic Index | 7.2 | 31.1 |
| AA Intelligence | 27 | 24 |
| AA-LCR | 66 | 63.3 |
| AA-Omniscience | -14.3 | -10.6 |
| aider_polyglot | 82.2 | 61.7 |
| AIME 2025 | 88.3 | 81.7 |
| AIR-Bench 2024 | 73.6 | 80 |
| anthropic_red_team | 99.5 | 98.3 |
| arena_vision | 1246 | 1193 |
| Artificial Analysis Coding Index | 46.7 | 39.7 |
| bbq | 96.4 | 97.3 |
| critpt | 2.6 | 0.3 |
| Fortress | 54.9 | 19.4 |
| gdpval | 8.5 | 11.5 |
| GPQA Diamond | 84.4 | 75.7 |
| harmbench | 65.4 | 96.3 |
| HLE | 22.5 | 7.7 |
| IFBench | 49 | 70.3 |
| MathVista | 83.9 | 71.8 |
| MMMU | 83.9 | 77.6 |
| OmniScience Accuracy | 39 | 34.7 |
| OmniScience Non-Hallucination | 12.6 | 30.7 |
| scicode | 43 | 35.8 |
| SEAL VISTA | 50.8 | 45.3 |
| simple_safety_tests | 97 | 99 |
| simplebench | 62.4 | 41.7 |
| simpleqa | 50.8 | 47 |
| SWE-bench Verified | 67.2 | 48.9 |
| Terminal-Bench Hard | 26.5 | 12.9 |
| xstest | 98.7 | 97 |
| τ²-Bench Telecom (AA run) | 54.1 | 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.