| Benchmark | Kimi K2.5 | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 52.8 | 8.8 |
| AA Intelligence | 36 | 25.7 |
| AA-LCR | 73 | 60.3 |
| AA-Omniscience | -7.3 | -41.5 |
| AIME 2025 | 96.1 | 90.2 |
| aime_2026 | 95.8 | 91.7 |
| apexAgents | 11.5 | 1.8 |
| Artificial Analysis Coding Index | 46.8 | 37.7 |
| browsecomp | 74.9 | 31.3 |
| critpt | 3.1 | 3.1 |
| gdpval | 38.3 | 9.9 |
| GPQA Diamond | 87.9 | 82.7 |
| HLE | 50.2 | 22.8 |
| HMMT 2025 | 95.4 | 94.7 |
| hmmt_feb_2026 | 87.1 | 85.2 |
| IFBench | 70.2 | 72.6 |
| MMLU-Pro | 87.1 | 83.8 |
| multichallenge | 61.4 | 55.2 |
| OmniScience Accuracy | 35.2 | 24.3 |
| OmniScience Non-Hallucination | 50 | 13 |
| scicode | 49 | 42 |
| SWE-bench Multilingual | 73 | 45.8 |
| SWE-bench Verified | 76.8 | 53.7 |
| TauBench V3 - Banking | 14.2 | 10.3 |
| Terminal-Bench 2.0 | 50.8 | 31 |
| Terminal-Bench 2.1 | 45.7 | 38.6 |
| Terminal-Bench Hard | 34.8 | 28.8 |
| τ²-Bench Telecom (AA run) | 95.9 | 67.8 |
| τ³-Bench | 66 | 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.