| Benchmark | Gemma 4 31B | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 14.4 | 8.8 |
| AA Intelligence | 29.7 | 25.7 |
| AA-LCR | 68.3 | 60.3 |
| AA-Omniscience | -47.9 | -41.5 |
| aime_2026 | 89.2 | 91.7 |
| Artificial Analysis Coding Index | 43.4 | 37.7 |
| critpt | 1.4 | 3.1 |
| gdpval | 15.5 | 9.9 |
| GPQA Diamond | 85.7 | 82.7 |
| HLE | 26.5 | 22.8 |
| hmmt_feb_2026 | 77.2 | 85.2 |
| IFBench | 75.6 | 72.6 |
| itbenchSre | 37.3 | 1.1 |
| MMLU-Pro | 85.2 | 83.8 |
| OmniScience Accuracy | 20 | 24.3 |
| OmniScience Non-Hallucination | 18.1 | 13 |
| scicode | 43.4 | 42 |
| SWE-bench Multilingual | 51.7 | 45.8 |
| SWE-bench Verified | 52 | 53.7 |
| TauBench V3 - Banking | 14.8 | 10.3 |
| Terminal-Bench 2.0 | 42.9 | 31 |
| Terminal-Bench 2.1 | 43.4 | 38.6 |
| Terminal-Bench Hard | 36.4 | 28.8 |
| τ²-Bench Telecom (AA run) | 65.5 | 67.8 |
| τ³-Bench | 67.5 | 10.1 |
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.