| Benchmark | GPT-5.6 Sol | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 57.8 | 47.9 |
| AA Intelligence | 61 | 33.5 |
| AA-LCR | 77.7 | 70.3 |
| AA-Omniscience | 22 | -21.4 |
| Artificial Analysis Coding Index | 78.3 | 34.8 |
| browsecomp | 90.4 | 60.2 |
| critpt | 32.3 | 2.6 |
| gdpval | 61.1 | 24.5 |
| GPQA Diamond | 94.6 | 84.5 |
| HealthBench | 57 | 58 |
| HLE | 49.5 | 23.9 |
| HLE (with tools) | 58 | 44.9 |
| IFBench | 72.7 | 68.1 |
| longbench_v2 | 67.1 | 45.1 |
| OmniScience Accuracy | 59.4 | 30.9 |
| OmniScience Non-Hallucination | 10.6 | 25.8 |
| ResearchRubrics | 73.8 | 56.2 |
| scicode | 56.9 | 44.8 |
| simplebench | 64.8 | 39.6 |
| Terminal-Bench Hard | 65.9 | 31.1 |
| τ²-Bench Telecom (AA run) | 85.1 | 93 |
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.