| Benchmark | Gemini 1.5 Pro | Gemini 3.1 Pro |
|---|---|---|
| AA Intelligence | 10 | 47.7 |
| ARC-AGI-2 | 0.8 | 77.1 |
| arena_vision | 1179 | 1280 |
| Artificial Analysis Coding Index | 23.6 | 68.8 |
| Chinese SimpleQA (C-SimpleQA) | 59.4 | 85.9 |
| Fortress | 53.9 | 29.8 |
| GPQA Diamond | 59.1 | 94.3 |
| HLE | 4.9 | 51.4 |
| LiveCodeBench | 30.5 | 91.7 |
| mmlu | 86.8 | 92.6 |
| MMLU-Pro | 75.8 | 91 |
| MMMU-Pro | 55 | 83 |
| scicode | 29.5 | 59 |
| simplebench | 27.1 | 79.6 |
| SWE-bench Verified | 34.2 | 80.6 |
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.