| Benchmark | Gemma 4 E2B | o1 |
|---|---|---|
| AA Agentic Index | 7.4 | 31.1 |
| AA Intelligence | 10 | 24 |
| AA-LCR | 17 | 63.3 |
| AA-Omniscience | -23.6 | -10.6 |
| Artificial Analysis Coding Index | 8.3 | 39.7 |
| critpt | 0 | 0.3 |
| gdpval | 0 | 11.5 |
| GPQA Diamond | 43.4 | 75.7 |
| HLE | 4.8 | 7.7 |
| IFBench | 38 | 70.3 |
| MathVista | 62.1 | 71.8 |
| mmmlu | 67.4 | 87.7 |
| OmniScience Accuracy | 8.2 | 34.7 |
| OmniScience Non-Hallucination | 67.6 | 30.7 |
| scicode | 20.9 | 35.8 |
| Terminal-Bench Hard | 3 | 12.9 |
| τ²-Bench Telecom (AA run) | 22.2 | 62.6 |
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.