| Benchmark | Claude 3.7 Sonnet | Gemma 4 E2B |
|---|---|---|
| AA Agentic Index | 37 | 7.4 |
| AA Intelligence | 27.6 | 10 |
| AA-LCR | 62.3 | 17 |
| AA-Omniscience | -0.7 | -23.6 |
| Artificial Analysis Coding Index | 36.4 | 8.3 |
| critpt | 0.9 | 0 |
| gdpval | 27.4 | 0 |
| GPQA Diamond | 84.8 | 43.4 |
| HLE | 9.7 | 4.8 |
| IFBench | 48.3 | 38 |
| ifeval | 93.2 | 82.9 |
| mmmlu | 86.1 | 67.4 |
| MMMU-Pro | 60.1 | 44.6 |
| OmniScience Accuracy | 28.2 | 8.2 |
| OmniScience Non-Hallucination | 60 | 67.6 |
| scicode | 40.3 | 20.9 |
| Terminal-Bench Hard | 21.2 | 3 |
| τ²-Bench Telecom (AA run) | 54.7 | 22.2 |
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.