| Benchmark | Gemma 4 E2B | Qwen3.6 Plus |
|---|---|---|
| AA Agentic Index | 7.4 | 29 |
| AA Intelligence | 10 | 40.5 |
| AA-LCR | 17 | 72.3 |
| AA-Omniscience | -23.6 | 2.6 |
| aime_2026 | 37.5 | 95.3 |
| Artificial Analysis Coding Index | 8.3 | 54.5 |
| critpt | 0 | 2.9 |
| gdpval | 0 | 32 |
| GPQA Diamond | 43.4 | 90.4 |
| HLE | 4.8 | 28.8 |
| IFBench | 38 | 75.2 |
| ifeval | 82.9 | 94.3 |
| LiveCodeBench v6 | 44 | 87.1 |
| mathvision | 52.4 | 88 |
| MMLU-Pro | 60 | 88.5 |
| mmmlu | 67.4 | 89.5 |
| MMMU-Pro | 44.6 | 78.8 |
| MMStar | 57.9 | 83.3 |
| OmniScience Accuracy | 8.2 | 26.4 |
| OmniScience Non-Hallucination | 67.6 | 68 |
| RealWorldQA | 56.2 | 85.4 |
| scicode | 20.9 | 40.7 |
| TauBench V3 - Banking | 4.5 | 20.8 |
| Terminal-Bench 2.1 | 0.4 | 61.4 |
| Terminal-Bench Hard | 3 | 43.9 |
| τ²-Bench Telecom (AA run) | 22.2 | 97.7 |
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.