| Benchmark | Claude 3.5 Sonnet | Gemma 4 E2B |
|---|---|---|
| AA Intelligence | 10 | 10 |
| Artificial Analysis Coding Index | 30.2 | 8.3 |
| BLINK | 56.5 | 51.8 |
| chartqa | 90.8 | 43.5 |
| GPQA Diamond | 67.2 | 43.4 |
| HLE | 3.9 | 4.8 |
| ifeval | 90.1 | 82.9 |
| LiveCodeBench v6 | 37.2 | 44 |
| MATH500 (Pass@1) | 78.3 | 64 |
| MathVista | 67.7 | 62.1 |
| MMLU-Pro | 78 | 60 |
| MMMU-Pro | 54.7 | 44.6 |
| MMStar | 62.2 | 57.9 |
| OCRBench | 790 | 0.7 |
| RealWorldQA | 60.1 | 56.2 |
| scicode | 36.6 | 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.