| Benchmark | Gemma 4 E2B | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 7.4 | 57.8 |
| AA Intelligence | 10 | 61 |
| AA-LCR | 17 | 77.7 |
| AA-Omniscience | -23.6 | 22 |
| aime_2026 | 37.5 | 99.9 |
| Artificial Analysis Coding Index | 8.3 | 78.3 |
| critpt | 0 | 32.3 |
| gdpval | 0 | 61.1 |
| GPQA Diamond | 43.4 | 94.6 |
| HLE | 4.8 | 49.5 |
| IFBench | 38 | 72.7 |
| mathvision | 52.4 | 95.8 |
| MMMU-Pro | 44.6 | 84.6 |
| OmniScience Accuracy | 8.2 | 59.4 |
| OmniScience Non-Hallucination | 67.6 | 10.6 |
| scicode | 20.9 | 56.9 |
| TauBench V3 - Banking | 4.5 | 44.3 |
| Terminal-Bench 2.1 | 0.4 | 89.5 |
| Terminal-Bench Hard | 3 | 65.9 |
| τ²-Bench Telecom (AA run) | 22.2 | 85.1 |
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.