| Benchmark | Kimi K2 (Reasoning) | Nemotron 3 Super 120B A12B |
|---|---|---|
| AA Agentic Index | 47.9 | 8.8 |
| AA Intelligence | 33.5 | 25.7 |
| AA-LCR | 70.3 | 60.3 |
| AA-Omniscience | -21.4 | -41.5 |
| AIME 2025 | 94.5 | 90.2 |
| AIME25 no tools | 94.5 | 92.2 |
| Artificial Analysis Coding Index | 34.8 | 37.7 |
| browsecomp | 60.2 | 31.3 |
| critpt | 2.6 | 3.1 |
| gdpval | 24.5 | 9.9 |
| GPQA Diamond | 84.5 | 82.7 |
| HLE | 23.9 | 22.8 |
| HMMT 2025 | 89.4 | 94.7 |
| IFBench | 68.1 | 72.6 |
| LiveCodeBench | 79.2 | 82.1 |
| MMLU-Pro | 84.6 | 83.8 |
| multichallenge | 66.4 | 55.2 |
| OmniScience Accuracy | 30.9 | 24.3 |
| OmniScience Non-Hallucination | 25.8 | 13 |
| scicode | 44.8 | 42 |
| SWE-bench Multilingual | 61.1 | 45.8 |
| SWE-bench Verified | 71.3 | 53.7 |
| Terminal-Bench | 47.1 | 25.8 |
| Terminal-Bench 2.0 | 35.7 | 31 |
| Terminal-Bench Hard | 31.1 | 28.8 |
| τ²-Bench Telecom (AA run) | 93 | 67.8 |
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.