| Benchmark | GLM-5.2 Full Open Source | GPT-4o |
|---|---|---|
| AA Agentic Index | 45.7 | 9.7 |
| AA Intelligence | 52.6 | 11.2 |
| AA-LCR | 76.7 | 55 |
| AA-Omniscience | 4.4 | -10.5 |
| arc_agi_1 | 77 | 4.5 |
| ARC-AGI-2 | 22.8 | 0 |
| Artificial Analysis Coding Index | 68.8 | 24.2 |
| critpt | 20.9 | 0 |
| gdpval | 50.3 | 0 |
| GPQA Diamond | 91.2 | 70.1 |
| HLE | 54.7 | 5.3 |
| IFBench | 73.3 | 36 |
| OmniScience Accuracy | 24.3 | 23.7 |
| OmniScience Non-Hallucination | 73.7 | 62.1 |
| scicode | 50.5 | 33.4 |
| simplebench | 58.8 | 17.8 |
| Terminal-Bench Hard | 50.8 | 8.3 |
| τ²-Bench Telecom (AA run) | 99.1 | 28.9 |
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.