| Benchmark | Claude 3.7 Sonnet | GLM-5.1 |
|---|---|---|
| AA Agentic Index | 37 | 66 |
| AA Intelligence | 27.6 | 41 |
| AA-LCR | 62.3 | 68 |
| AA-Omniscience | -0.7 | 0.8 |
| Artificial Analysis Coding Index | 36.4 | 55.8 |
| critpt | 0.9 | 4.6 |
| gdpval | 27.4 | 49.5 |
| GPQA Diamond | 84.8 | 86.8 |
| HLE | 9.7 | 52.3 |
| IFBench | 48.3 | 76.3 |
| OmniScience Accuracy | 28.2 | 25.2 |
| OmniScience Non-Hallucination | 60 | 70.1 |
| scicode | 40.3 | 43.8 |
| simplebench | 46.4 | 55.1 |
| Terminal-Bench Hard | 21.2 | 43.2 |
| τ²-Bench Telecom (AA run) | 54.7 | 97.7 |
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.