| Benchmark | GPT-5.4 | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 58.2 | 8.8 |
| AA Intelligence | 53.1 | 25.7 |
| AA-LCR | 77.7 | 60.3 |
| AA-Omniscience | 5.8 | -41.5 |
| aime_2026 | 99.2 | 91.7 |
| apexAgents | 33.3 | 1.8 |
| Artificial Analysis Coding Index | 71.1 | 37.7 |
| browsecomp | 82.7 | 31.3 |
| critpt | 23.4 | 3.1 |
| gdpval | 50.1 | 9.9 |
| GPQA Diamond | 93 | 82.7 |
| HLE | 43.7 | 22.8 |
| hmmt_feb_2026 | 97.7 | 85.2 |
| IFBench | 73.9 | 72.6 |
| itbenchSre | 34.5 | 1.1 |
| MMLU-Pro | 87.5 | 83.8 |
| OmniScience Accuracy | 50.9 | 24.3 |
| OmniScience Non-Hallucination | 17.4 | 13 |
| scicode | 56.6 | 42 |
| SWE-bench Multilingual | 71.7 | 45.8 |
| TauBench V3 - Banking | 39.6 | 10.3 |
| Terminal-Bench 2.0 | 75.1 | 31 |
| Terminal-Bench 2.1 | 78.3 | 38.6 |
| Terminal-Bench Hard | 57.6 | 28.8 |
| τ²-Bench Telecom (AA run) | 87.1 | 67.8 |
| τ³-Bench | 72.9 | 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.