| Benchmark | Nemotron 3 Super 120B A12B | Qwen3.5 397B A17B |
|---|---|---|
| AA Agentic Index | 8.8 | 53.3 |
| AA Intelligence | 25.7 | 34.3 |
| AA-LCR | 60.3 | 72.7 |
| AA-Omniscience | -41.5 | -30.8 |
| aime_2026 | 91.7 | 94.2 |
| apexAgents | 1.8 | 15.3 |
| Artificial Analysis Coding Index | 37.7 | 48.2 |
| browsecomp | 31.3 | 69 |
| critpt | 3.1 | 1.7 |
| gdpval | 9.9 | 35.8 |
| GPQA Diamond | 82.7 | 89.3 |
| GSM8K | 90.7 | 66.7 |
| HLE | 22.8 | 29 |
| HMMT 2025 | 94.7 | 92.7 |
| hmmt_feb_2026 | 85.2 | 87.9 |
| IFBench | 72.6 | 78.8 |
| itbenchSre | 1.1 | 34.1 |
| mmlu_prox | 79.4 | 84.7 |
| MMLU-Pro | 83.8 | 87.8 |
| multichallenge | 55.2 | 67.6 |
| OmniScience Accuracy | 24.3 | 30.8 |
| OmniScience Non-Hallucination | 13 | 17.3 |
| scicode | 42 | 42 |
| SWE-bench Multilingual | 45.8 | 69.3 |
| SWE-bench Verified | 53.7 | 76.4 |
| TauBench V3 - Banking | 10.3 | 13.4 |
| Terminal-Bench 2.0 | 31 | 52.5 |
| Terminal-Bench 2.1 | 38.6 | 51.3 |
| Terminal-Bench Hard | 28.8 | 40.9 |
| τ²-Bench Telecom (AA run) | 67.8 | 95.6 |
| τ³-Bench | 10.1 | 13.4 |
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.