| Benchmark | DeepSeek-V3 | Gemma 4 E2B |
|---|---|---|
| AA Agentic Index | 1.6 | 7.4 |
| AA Intelligence | 15.4 | 10 |
| AA-LCR | 41.3 | 17 |
| AA-Omniscience | -37.6 | -23.6 |
| Artificial Analysis Coding Index | 23 | 8.3 |
| critpt | 0 | 0 |
| gdpval | 0 | 0 |
| GPQA Diamond | 68.4 | 43.4 |
| HLE | 5.2 | 4.8 |
| IFBench | 41 | 38 |
| ifeval | 87.3 | 82.9 |
| LiveCodeBench v6 | 46.9 | 44 |
| MMLU-Pro | 81.2 | 60 |
| mmmlu | 79.4 | 67.4 |
| OmniScience Accuracy | 25.4 | 8.2 |
| OmniScience Non-Hallucination | 23.3 | 67.6 |
| scicode | 35.8 | 20.9 |
| Tau2 retail | 69.1 | 18.9 |
| TauBench V3 - Banking | 4.7 | 4.5 |
| Terminal-Bench 2.1 | 16.9 | 0.4 |
| Terminal-Bench Hard | 15.2 | 3 |
| τ²-Bench | 32.5 | 24.5 |
| τ²-Bench Telecom (AA run) | 47.1 | 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.