| Benchmark | Claude Fable 5 | Qwen3.6 35B A3B |
|---|---|---|
| AA Agentic Index | 56.6 | 21.6 |
| AA Intelligence | 62.1 | 32.1 |
| AA-LCR | 76.7 | 66.7 |
| AA-Omniscience | 43.3 | -22.2 |
| Artificial Analysis Coding Index | 76.5 | 41.9 |
| browsecomp | 88 | 48.7 |
| charxiv_rq | 89.4 | 78 |
| critpt | 28.6 | 0.3 |
| gdpval | 61.9 | 27.8 |
| GDPval-AA v2 | 1760 | 1015 |
| GPQA Diamond | 92.6 | 86 |
| HLE | 64.5 | 22.2 |
| IFBench | 63.5 | 64.4 |
| MCP Atlas | 84.7 | 62.8 |
| MMMU-Pro | 84.2 | 75.3 |
| OmniScience Accuracy | 65.3 | 18.8 |
| OmniScience Non-Hallucination | 36.4 | 49.5 |
| scicode | 60.2 | 35.8 |
| SWE-bench Multilingual | 86.6 | 67.2 |
| SWE-bench Pro | 80 | 49.5 |
| SWE-bench Verified | 95 | 73.4 |
| TauBench V3 - Banking | 38.1 | 9.3 |
| Terminal-Bench 2.1 | 88 | 44.9 |
| Terminal-Bench Hard | 62.9 | 34.8 |
| toolathlon | 61.7 | 26.9 |
| WideSearch | 81.2 | 60.1 |
| τ²-Bench Telecom (AA run) | 98.5 | 95.3 |
| τ³-Bench | 26.8 | 67.2 |
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.