| Benchmark | GPT-5.2 | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 60.2 | 8.8 |
| AA Intelligence | 43.3 | 25.7 |
| AA-LCR | 79.3 | 60.3 |
| AA-Omniscience | 0.3 | -41.5 |
| AIME 2025 | 100 | 90.2 |
| aime_2026 | 98.3 | 91.7 |
| AIME25 no tools | 98 | 92.2 |
| apexAgents | 23 | 1.8 |
| Artificial Analysis Coding Index | 48.7 | 37.7 |
| browsecomp | 65.8 | 31.3 |
| critpt | 11.6 | 3.1 |
| gdpval | 48.3 | 9.9 |
| GPQA Diamond | 92.4 | 82.7 |
| HLE | 45.5 | 22.8 |
| HMMT 2025 | 99.4 | 94.7 |
| hmmt_feb_2026 | 97 | 85.2 |
| IFBench | 75.4 | 72.6 |
| LiveCodeBench | 89 | 82.1 |
| MMLU-Pro | 87 | 83.8 |
| OmniScience Accuracy | 44.3 | 24.3 |
| OmniScience Non-Hallucination | 38.4 | 13 |
| scicode | 52.1 | 42 |
| SWE-bench Multilingual | 72 | 45.8 |
| SWE-bench Verified | 80 | 53.7 |
| Tau2 retail | 82 | 62.8 |
| TauBench V3 - Banking | 11.1 | 10.3 |
| Terminal-Bench 2.0 | 54 | 31 |
| Terminal-Bench Hard | 62.2 | 28.8 |
| τ²-Bench Telecom (AA run) | 98.7 | 67.8 |
| τ³-Bench | 12 | 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.