| Benchmark | Claude Sonnet 4 | GLM-5.1 |
|---|---|---|
| AA Agentic Index | 39.2 | 66 |
| AA Intelligence | 29.8 | 41 |
| AA-LCR | 69.7 | 68 |
| AA-Omniscience | 0.2 | 0.8 |
| Artificial Analysis Coding Index | 37.6 | 55.8 |
| browsecomp | 12.2 | 79.3 |
| critpt | 1.1 | 4.6 |
| frontiermath_tier_4 | 0 | 12.5 |
| gdpval | 31.2 | 49.5 |
| GPQA Diamond | 78 | 86.8 |
| HLE | 10.7 | 52.3 |
| HLE (with tools) | 20.3 | 52.3 |
| HMMT 2025 | 15.9 | 94 |
| IFBench | 55 | 76.3 |
| livebench | 74.8 | 70.2 |
| MMLU-Pro | 84 | 86 |
| OmniScience Accuracy | 22.7 | 25.2 |
| OmniScience Non-Hallucination | 70.9 | 70.1 |
| scicode | 40 | 43.8 |
| simplebench | 45.5 | 55.1 |
| SWE-bench Multilingual | 56.9 | 73.3 |
| SWE-bench Pro | 42.7 | 58.4 |
| TauBench V3 - Banking | 16.7 | 13.6 |
| Terminal-Bench 2.1 | 36.3 | 63.5 |
| Terminal-Bench Hard | 31.1 | 43.2 |
| τ²-Bench Telecom (AA run) | 64.6 | 97.7 |
| τ³-Bench | 13.8 | 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.