| Benchmark | Claude Sonnet 4.6 | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 61.6 | 8.8 |
| AA Intelligence | 48.4 | 25.7 |
| AA-LCR | 74 | 60.3 |
| AA-Omniscience | 12.2 | -41.5 |
| apexAgents | 28 | 1.8 |
| Artificial Analysis Coding Index | 63 | 37.7 |
| browsecomp | 76.2 | 31.3 |
| critpt | 3.1 | 3.1 |
| gdpval | 54.8 | 9.9 |
| GPQA Diamond | 89.9 | 82.7 |
| HLE | 49 | 22.8 |
| IFBench | 56.6 | 72.6 |
| itbenchSre | 39.8 | 1.1 |
| OmniScience Accuracy | 40.9 | 24.3 |
| OmniScience Non-Hallucination | 51.6 | 13 |
| scicode | 47 | 42 |
| SWE-bench Verified | 79.6 | 53.7 |
| Tau2 retail | 91.7 | 62.8 |
| TauBench V3 - Banking | 34.4 | 10.3 |
| Terminal-Bench 2.0 | 59.1 | 31 |
| Terminal-Bench 2.1 | 71.2 | 38.6 |
| Terminal-Bench Hard | 59.1 | 28.8 |
| τ²-Bench Telecom (AA run) | 97.9 | 67.8 |
| τ³-Bench | 30.5 | 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.