| Benchmark | Gemma 4 E2B | GPT-4o mini |
|---|---|---|
| AA Agentic Index | 7.4 | 1 |
| AA Intelligence | 10 | 7 |
| Artificial Analysis Coding Index | 8.3 | 11.4 |
| BLINK | 51.8 | 51.9 |
| gdpval | 0 | 0 |
| GPQA Diamond | 43.4 | 42.6 |
| HLE | 4.8 | 4.2 |
| IFBench | 38 | 31 |
| MathVista | 62.1 | 56.7 |
| MMMU-Pro | 44.6 | 41.5 |
| scicode | 20.9 | 22.9 |
| TauBench V3 - Banking | 4.5 | 2.9 |
| Terminal-Bench 2.1 | 0.4 | 5.6 |
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.