| Benchmark | Claude Fable 5 | GLM-5.1 |
|---|---|---|
| AA Agentic Index | 56.6 | 66 |
| AA Intelligence | 62.1 | 41 |
| AA-LCR | 76.7 | 68 |
| AA-Omniscience | 43.3 | 0.8 |
| arena_elo | 1508 | 1468 |
| arena_text_factuality | 1486 | 1458 |
| Artificial Analysis Coding Index | 76.5 | 55.8 |
| browsecomp | 88 | 79.3 |
| coding_arena_elo | 1626 | 1510 |
| critpt | 28.6 | 4.6 |
| DeepSWE | 70 | 18 |
| Finance Agent v2 | 56.3 | 44.8 |
| gdpval | 61.9 | 49.5 |
| GDPval-AA (Elo) | 1932 | 1535 |
| GPQA Diamond | 92.6 | 86.8 |
| HLE | 64.5 | 52.3 |
| HLE (with tools) | 64.5 | 52.3 |
| IFBench | 63.5 | 76.3 |
| livebench | 78.3 | 70.2 |
| MCP Atlas | 84.7 | 75.6 |
| OmniScience Accuracy | 65.3 | 25.2 |
| OmniScience Non-Hallucination | 36.4 | 70.1 |
| PostTrainBench | 41.4 | 20.1 |
| Program Bench | 76.8 | 50.9 |
| scicode | 60.2 | 43.8 |
| simplebench | 81.9 | 55.1 |
| simpleqa_verified | 68.3 | 38.1 |
| SWE-bench Multilingual | 86.6 | 73.3 |
| SWE-bench Pro | 80 | 58.4 |
| SWE-Marathon | 35 | 1 |
| TauBench V3 - Banking | 38.1 | 13.6 |
| Terminal-Bench 2.1 | 88 | 63.5 |
| Terminal-Bench Hard | 62.9 | 43.2 |
| toolathlon | 61.7 | 40.7 |
| τ²-Bench Telecom (AA run) | 98.5 | 97.7 |
| τ³-Bench | 26.8 | 70.6 |
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.