| Benchmark | Gemini 2.5 Flash | o1 |
|---|---|---|
| AA Agentic Index | 18.8 | 31.1 |
| AA Intelligence | 17.7 | 24 |
| AA-LCR | 65.7 | 63.3 |
| AA-Omniscience | -29.8 | -10.6 |
| aider_polyglot | 61.9 | 61.7 |
| AIME 2024 | 82.3 | 79.2 |
| AIME 2025 | 72 | 81.7 |
| AIR-Bench 2024 | 68.7 | 80 |
| anthropic_red_team | 98.8 | 98.3 |
| arena_vision | 1214 | 1193 |
| Artificial Analysis Coding Index | 22.2 | 39.7 |
| bbq | 97.7 | 97.3 |
| critpt | 1.4 | 0.3 |
| gdpval | 11.9 | 11.5 |
| GPQA Diamond | 82.8 | 75.7 |
| harmbench | 62.6 | 96.3 |
| HLE | 12.1 | 7.7 |
| IFBench | 50.3 | 70.3 |
| MMMU | 79.7 | 77.6 |
| OmniScience Accuracy | 26.1 | 34.7 |
| OmniScience Non-Hallucination | 24.7 | 30.7 |
| scicode | 35.9 | 35.8 |
| SEAL VISTA | 49.1 | 45.3 |
| simple_safety_tests | 98 | 99 |
| simplebench | 41.2 | 41.7 |
| simpleqa | 26.9 | 47 |
| SWE-bench Verified | 60.4 | 48.9 |
| Terminal-Bench Hard | 13.6 | 12.9 |
| xstest | 98.8 | 97 |
| τ²-Bench Telecom (AA run) | 31.6 | 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.