| Benchmark | Claude 3.5 Sonnet | Gemini 3 Flash Preview |
|---|---|---|
| AA Intelligence | 10 | 38.7 |
| AIME 2025 | 7.4 | 99.7 |
| arena_vision | 1146 | 1260 |
| Artificial Analysis Coding Index | 30.2 | 42.6 |
| frontiermath_tier_4 | 0 | 4.2 |
| GPQA Diamond | 67.2 | 90.4 |
| harmbench | 98.1 | 70 |
| HLE | 3.9 | 43.5 |
| hmmt_feb_2025 | 1.7 | 97.5 |
| MMMU-Pro | 54.7 | 81.2 |
| scicode | 36.6 | 50.6 |
| simplebench | 41.4 | 61.1 |
| simpleqa | 28.4 | 68.7 |
| SWE-bench Verified | 50.8 | 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.