| Benchmark | Claude Sonnet 4 | Gemma 4 E2B |
|---|---|---|
| AA Agentic Index | 39.2 | 7.4 |
| AA Intelligence | 29.8 | 10 |
| AA-LCR | 69.7 | 17 |
| AA-Omniscience | 0.2 | -23.6 |
| Artificial Analysis Coding Index | 37.6 | 8.3 |
| critpt | 1.1 | 0 |
| gdpval | 31.2 | 0 |
| GPQA Diamond | 78 | 43.4 |
| HLE | 10.7 | 4.8 |
| IFBench | 55 | 38 |
| ifeval | 87.6 | 82.9 |
| LiveCodeBench v6 | 48.5 | 44 |
| MMLU-Pro | 84 | 60 |
| mmmlu | 86.5 | 67.4 |
| MMMU-Pro | 62.4 | 44.6 |
| OmniScience Accuracy | 22.7 | 8.2 |
| OmniScience Non-Hallucination | 70.9 | 67.6 |
| scicode | 40 | 20.9 |
| Tau2 retail | 75 | 18.9 |
| TauBench V3 - Banking | 16.7 | 4.5 |
| Terminal-Bench 2.1 | 36.3 | 0.4 |
| Terminal-Bench Hard | 31.1 | 3 |
| τ²-Bench | 65 | 24.5 |
| τ²-Bench Telecom (AA run) | 64.6 | 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.