| Benchmark | Claude Fable 5 | Qwen3.6 Plus |
|---|---|---|
| AA Agentic Index | 56.6 | 29 |
| AA Intelligence | 62.1 | 40.5 |
| AA-LCR | 76.7 | 72.3 |
| AA-Omniscience | 43.3 | 2.6 |
| arena_elo | 1508 | 1444 |
| arena_text_factuality | 1486 | 1438 |
| Artificial Analysis Coding Index | 76.5 | 54.5 |
| coding_arena_elo | 1626 | 1460 |
| critpt | 28.6 | 2.9 |
| Finance Agent v2 | 56.3 | 40.9 |
| gdpval | 61.9 | 32 |
| GPQA Diamond | 92.6 | 90.4 |
| HLE | 64.5 | 28.8 |
| HLE (with tools) | 64.5 | 50.6 |
| IFBench | 63.5 | 75.2 |
| livebench | 78.3 | 70.8 |
| livebench_agentic_coding | 62.2 | 41.4 |
| livebench_coding | 86 | 78.2 |
| livebench_data_analysis | 80.5 | 69.9 |
| livebench_instruction_following | 75.8 | 58.3 |
| livebench_language | 90.7 | 75 |
| livebench_math | 96 | 83.7 |
| livebench_reasoning | 89.7 | 75.8 |
| mathvision | 94.8 | 88 |
| MCP Atlas | 84.7 | 74.1 |
| MMMU-Pro | 84.2 | 78.8 |
| OmniScience Accuracy | 65.3 | 26.4 |
| OmniScience Non-Hallucination | 36.4 | 68 |
| OSWorld-Verified | 85 | 62.5 |
| scicode | 60.2 | 40.7 |
| simpleqa_verified | 68.3 | 49.1 |
| SWE-bench Multilingual | 86.6 | 73.8 |
| SWE-bench Pro | 80 | 56.6 |
| SWE-bench Verified | 95 | 78.8 |
| TauBench V3 - Banking | 38.1 | 20.8 |
| Terminal-Bench 2.1 | 88 | 61.4 |
| Terminal-Bench Hard | 62.9 | 43.9 |
| toolathlon | 61.7 | 39.8 |
| WideSearch | 81.2 | 74.3 |
| τ²-Bench Telecom (AA run) | 98.5 | 97.7 |
| τ³-Bench | 26.8 | 70.7 |
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.