| Benchmark | Gemma 4 31B | GPT-4o mini |
|---|---|---|
| AA Agentic Index | 14.4 | 1 |
| AA Intelligence | 29.7 | 7 |
| AI2D | 89 | 75.2 |
| Artificial Analysis Coding Index | 43.4 | 11.4 |
| gdpval | 15.5 | 0 |
| GPQA Diamond | 85.7 | 42.6 |
| HLE | 26.5 | 4.2 |
| IFBench | 75.6 | 31 |
| MathVista | 79.3 | 56.7 |
| MMMU | 80.4 | 59.4 |
| MMMU-Pro | 76.9 | 41.5 |
| scicode | 43.4 | 22.9 |
| SWE-bench Verified | 52 | 8.7 |
| TauBench V3 - Banking | 14.8 | 2.9 |
| Terminal-Bench 2.1 | 43.4 | 5.6 |
| τ³-Bench | 67.5 | 2.9 |
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.