| Benchmark | Gemini 3 Pro | GPT-4o mini |
|---|---|---|
| AA Agentic Index | 52 | 1 |
| AA Intelligence | 40.6 | 7 |
| ARC-AGI-2 | 54 | 0 |
| Artificial Analysis Coding Index | 46.5 | 11.4 |
| Fortress | 41.7 | 48.1 |
| gdpval | 34.2 | 0 |
| GPQA Diamond | 91.9 | 42.6 |
| HLE | 45.8 | 4.2 |
| IFBench | 70.4 | 31 |
| LiveCodeBench | 92 | 35.5 |
| MathVista | 89.8 | 56.7 |
| MMMU-Pro | 81 | 41.5 |
| scicode | 56.1 | 22.9 |
| simplebench | 76.4 | 10.7 |
| simpleqa | 72.1 | 9.9 |
| SWE-bench Verified | 78 | 8.7 |
| Video-MME | 88.4 | 68.9 |
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.