| Benchmark | GPT-5.6 Sol | Qwen3.6 35B A3B |
|---|---|---|
| AA Agentic Index | 57.8 | 21.6 |
| AA Intelligence | 61 | 32.1 |
| AA-LCR | 77.7 | 66.7 |
| AA-Omniscience | 22 | -22.2 |
| aime_2026 | 99.9 | 92.7 |
| Artificial Analysis Coding Index | 78.3 | 41.9 |
| browsecomp | 90.4 | 48.7 |
| charxiv_rq | 87.8 | 78 |
| critpt | 32.3 | 0.3 |
| gdpval | 61.1 | 27.8 |
| GDPval-AA v2 | 1748 | 1015 |
| GPQA Diamond | 94.6 | 86 |
| HLE | 49.5 | 22.2 |
| IFBench | 72.7 | 64.4 |
| MCP Atlas | 83.6 | 62.8 |
| MMMU-Pro | 84.6 | 75.3 |
| OmniScience Accuracy | 59.4 | 18.8 |
| OmniScience Non-Hallucination | 10.6 | 49.5 |
| scicode | 56.9 | 35.8 |
| SWE-bench Pro | 64.6 | 49.5 |
| TauBench V3 - Banking | 44.3 | 9.3 |
| Terminal-Bench 2.1 | 89.5 | 44.9 |
| Terminal-Bench Hard | 65.9 | 34.8 |
| toolathlon | 58 | 26.9 |
| τ²-Bench Telecom (AA run) | 85.1 | 95.3 |
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.