| Benchmark | Gemini 3.1 Pro | o1 |
|---|---|---|
| AA Agentic Index | 23 | 31.1 |
| AA Intelligence | 47.7 | 24 |
| AA-LCR | 79 | 63.3 |
| AA-Omniscience | 32.9 | -10.6 |
| arena_vision | 1280 | 1193 |
| Artificial Analysis Coding Index | 68.8 | 39.7 |
| Codeforces | 3052 | 2061 |
| critpt | 17.7 | 0.3 |
| Fortress | 29.8 | 19.4 |
| gdpval | 23.3 | 11.5 |
| GPQA Diamond | 94.3 | 75.7 |
| HLE | 51.4 | 7.7 |
| IFBench | 77.1 | 70.3 |
| livebench | 79.9 | 52.3 |
| mmlu | 92.6 | 91.8 |
| mmmlu | 92.6 | 87.7 |
| OmniScience Accuracy | 55.3 | 34.7 |
| OmniScience Non-Hallucination | 50.1 | 30.7 |
| scicode | 59 | 35.8 |
| simplebench | 79.6 | 41.7 |
| strongreject | 98 | 97 |
| SWE-bench Verified | 80.6 | 48.9 |
| Terminal-Bench Hard | 68.5 | 12.9 |
| τ²-Bench Telecom (AA run) | 99.3 | 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.