| Benchmark | GLM-5.2 Full Open Source | o4-mini |
|---|---|---|
| AA Agentic Index | 45.7 | 36.1 |
| AA Intelligence | 52.6 | 26.1 |
| AA-LCR | 76.7 | 60 |
| AA-Omniscience | 4.4 | -35.7 |
| arc_agi_1 | 77 | 58.7 |
| ARC-AGI-2 | 22.8 | 6.1 |
| Artificial Analysis Coding Index | 68.8 | 25.6 |
| critpt | 20.9 | 0.6 |
| gdpval | 50.3 | 25.4 |
| GPQA Diamond | 91.2 | 81.4 |
| HLE | 54.7 | 16.5 |
| IFBench | 73.3 | 68.7 |
| OmniScience Accuracy | 24.3 | 24.8 |
| OmniScience Non-Hallucination | 73.7 | 19.5 |
| scicode | 50.5 | 46.5 |
| simplebench | 58.8 | 38.7 |
| simpleqa_verified | 38.1 | 23.9 |
| strongreject | 98.5 | 96 |
| Terminal-Bench Hard | 50.8 | 15.2 |
| τ²-Bench Telecom (AA run) | 99.1 | 55.6 |
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.