| Benchmark | Gemini 1.5 Pro | Gemma 4 E2B |
|---|---|---|
| AA Intelligence | 10 | 10 |
| Artificial Analysis Coding Index | 23.6 | 8.3 |
| BLINK | 61 | 51.8 |
| chartqa | 88.7 | 43.5 |
| GPQA Diamond | 59.1 | 43.4 |
| HLE | 4.9 | 4.8 |
| ifeval | 89.4 | 82.9 |
| MathVista | 70.6 | 62.1 |
| MMLU-Pro | 75.8 | 60 |
| MMMU-Pro | 55 | 44.6 |
| OCRBench | 800 | 0.7 |
| ocrbench_v2 | 51.6 | 44.7 |
| scicode | 29.5 | 20.9 |
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.