| Benchmark | Gemma 4 E2B | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 7.4 | 20.8 |
| AA Intelligence | 10 | 39.4 |
| AA-LCR | 17 | 69 |
| AA-Omniscience | -23.6 | 2.4 |
| Artificial Analysis Coding Index | 8.3 | 55.9 |
| bfcl_v4 | 37 | 72.9 |
| critpt | 0 | 9.1 |
| gdpval | 0 | 22.2 |
| GPQA Diamond | 43.4 | 90.3 |
| HLE | 4.8 | 35.6 |
| IFBench | 38 | 79.1 |
| ifeval | 82.9 | 94.6 |
| LiveCodeBench v6 | 44 | 89.6 |
| mathvision | 52.4 | 90.3 |
| MedXPertQA MM | 23.5 | 71 |
| MMLU-Pro | 60 | 88.5 |
| mmmlu | 67.4 | 89 |
| MMMU-Pro | 44.6 | 80.5 |
| ocrbench_v2 | 44.7 | 67.1 |
| OmniScience Accuracy | 8.2 | 22.5 |
| OmniScience Non-Hallucination | 67.6 | 74.5 |
| RealWorldQA | 56.2 | 86.9 |
| scicode | 20.9 | 51.3 |
| TauBench V3 - Banking | 4.5 | 17.9 |
| Terminal-Bench 2.1 | 0.4 | 61 |
| Terminal-Bench Hard | 3 | 47 |
| τ²-Bench Telecom (AA run) | 22.2 | 93 |
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.