| Benchmark | Claude Opus 4.8 | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 49.4 | 8.8 |
| AA Intelligence | 57.3 | 25.7 |
| AA-LCR | 73 | 60.3 |
| AA-Omniscience | 28.8 | -41.5 |
| aime_2026 | 100 | 91.7 |
| Artificial Analysis Coding Index | 74.3 | 37.7 |
| browsecomp | 88.5 | 31.3 |
| critpt | 20.9 | 3.1 |
| gdpval | 54.2 | 9.9 |
| GPQA Diamond | 93.6 | 82.7 |
| HLE | 57.9 | 22.8 |
| hmmt_feb_2026 | 96.7 | 85.2 |
| IFBench | 62.2 | 72.6 |
| OmniScience Accuracy | 48.8 | 24.3 |
| OmniScience Non-Hallucination | 60.7 | 13 |
| scicode | 53.5 | 42 |
| SWE-bench Multilingual | 84.4 | 45.8 |
| SWE-bench Verified | 88.6 | 53.7 |
| TauBench V3 - Banking | 34.2 | 10.3 |
| Terminal-Bench 2.0 | 74.6 | 31 |
| Terminal-Bench 2.1 | 85 | 38.6 |
| Terminal-Bench Hard | 58.3 | 28.8 |
| τ²-Bench Telecom (AA run) | 94.4 | 67.8 |
| τ³-Bench | 27.6 | 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.