| Benchmark | Gemini 1.5 Pro | Gemini 3 Flash Preview |
|---|---|---|
| AA Intelligence | 10 | 38.7 |
| ARC-AGI-2 | 0.8 | 33.6 |
| arena_vision | 1179 | 1260 |
| Artificial Analysis Coding Index | 23.6 | 42.6 |
| GPQA Diamond | 59.1 | 90.4 |
| harmbench | 79.9 | 70 |
| HLE | 4.9 | 43.5 |
| MMMU-Pro | 55 | 81.2 |
| scicode | 29.5 | 50.6 |
| simplebench | 27.1 | 61.1 |
| simpleqa | 23.4 | 68.7 |
| SWE-bench Verified | 34.2 | 78 |
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.