| Benchmark | GLM-5.2 Full Open Source | o1 |
|---|---|---|
| AA Agentic Index | 45.7 | 31.1 |
| AA Intelligence | 52.6 | 24 |
| AA-LCR | 76.7 | 63.3 |
| AA-Omniscience | 4.4 | -10.6 |
| Artificial Analysis Coding Index | 68.8 | 39.7 |
| critpt | 20.9 | 0.3 |
| gdpval | 50.3 | 11.5 |
| GPQA Diamond | 91.2 | 75.7 |
| HLE | 54.7 | 7.7 |
| IFBench | 73.3 | 70.3 |
| OmniScience Accuracy | 24.3 | 34.7 |
| OmniScience Non-Hallucination | 73.7 | 30.7 |
| scicode | 50.5 | 35.8 |
| simplebench | 58.8 | 41.7 |
| strongreject | 98.5 | 97 |
| Terminal-Bench Hard | 50.8 | 12.9 |
| τ²-Bench Telecom (AA run) | 99.1 | 62.6 |
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.