| Benchmark | Gemini 2.5 Pro | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 7.2 | 8.8 |
| AA Intelligence | 27 | 25.7 |
| AA-LCR | 66 | 60.3 |
| AA-Omniscience | -14.3 | -41.5 |
| AIME 2025 | 88.3 | 90.2 |
| Artificial Analysis Coding Index | 46.7 | 37.7 |
| browsecomp | 9.9 | 31.3 |
| critpt | 2.6 | 3.1 |
| gdpval | 8.5 | 9.9 |
| GPQA Diamond | 84.4 | 82.7 |
| HLE | 22.5 | 22.8 |
| HMMT 2025 | 65.7 | 94.7 |
| IFBench | 49 | 72.6 |
| LiveCodeBench | 82.7 | 82.1 |
| MMLU-Pro | 86 | 83.8 |
| multichallenge | 53.6 | 55.2 |
| OmniScience Accuracy | 39 | 24.3 |
| OmniScience Non-Hallucination | 12.6 | 13 |
| scicode | 43 | 42 |
| SWE-bench Verified | 67.2 | 53.7 |
| TauBench V3 - Banking | 9.7 | 10.3 |
| Terminal-Bench | 25.3 | 25.8 |
| Terminal-Bench 2.1 | 28.5 | 38.6 |
| Terminal-Bench Hard | 26.5 | 28.8 |
| τ²-Bench Telecom (AA run) | 54.1 | 67.8 |
| τ³-Bench | 9.3 | 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.