| Benchmark | GPT-5.5 | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 47.4 | 8.8 |
| AA Intelligence | 56.3 | 25.7 |
| AA-LCR | 79 | 60.3 |
| AA-Omniscience | 20.5 | -41.5 |
| aime_2026 | 100 | 91.7 |
| apexAgents | 37.7 | 1.8 |
| Artificial Analysis Coding Index | 74.9 | 37.7 |
| browsecomp | 84.4 | 31.3 |
| critpt | 27.1 | 3.1 |
| gdpval | 49.5 | 9.9 |
| GPQA Diamond | 93.6 | 82.7 |
| HLE | 52.2 | 22.8 |
| hmmt_feb_2026 | 98.5 | 85.2 |
| IFBench | 75.9 | 72.6 |
| itbenchSre | 45.8 | 1.1 |
| OmniScience Accuracy | 58 | 24.3 |
| OmniScience Non-Hallucination | 12 | 13 |
| scicode | 56.1 | 42 |
| SWE-bench Verified | 80.6 | 53.7 |
| TauBench V3 - Banking | 39 | 10.3 |
| Terminal-Bench 2.0 | 82.7 | 31 |
| Terminal-Bench 2.1 | 84.3 | 38.6 |
| Terminal-Bench Hard | 60.6 | 28.8 |
| τ²-Bench Telecom (AA run) | 93.9 | 67.8 |
| τ³-Bench | 31.3 | 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.