| Benchmark | Gemma 4 E2B | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 7.4 | 8.8 |
| AA Intelligence | 10 | 25.7 |
| AA-LCR | 17 | 60.3 |
| AA-Omniscience | -23.6 | -41.5 |
| aime_2026 | 37.5 | 91.7 |
| AIME25 no tools | 26.3 | 92.2 |
| Artificial Analysis Coding Index | 8.3 | 37.7 |
| critpt | 0 | 3.1 |
| gdpval | 0 | 9.9 |
| GPQA Diamond | 43.4 | 82.7 |
| HLE | 4.8 | 22.8 |
| IFBench | 38 | 72.6 |
| MMLU-Pro | 60 | 83.8 |
| OmniScience Accuracy | 8.2 | 24.3 |
| OmniScience Non-Hallucination | 67.6 | 13 |
| scicode | 20.9 | 42 |
| Tau2 retail | 18.9 | 62.8 |
| TauBench V3 - Banking | 4.5 | 10.3 |
| Terminal-Bench 2.1 | 0.4 | 38.6 |
| Terminal-Bench Hard | 3 | 28.8 |
| τ²-Bench Telecom (AA run) | 22.2 | 67.8 |
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.