| Benchmark | Claude 4 Opus | GLM-5.2 Full Open Source |
|---|---|---|
| AA Intelligence | 31.7 | 52.6 |
| AA-LCR | 40 | 76.7 |
| arc_agi_1 | 27 | 77 |
| ARC-AGI-2 | 8.6 | 22.8 |
| Artificial Analysis Coding Index | 34 | 68.8 |
| GPQA Diamond | 79.6 | 91.2 |
| HLE | 12.3 | 54.7 |
| HMMT 2025 | 15.9 | 94.4 |
| IFBench | 53.7 | 73.3 |
| scicode | 40.9 | 50.5 |
| simplebench | 58.8 | 58.8 |
| Terminal-Bench Hard | 31.1 | 50.8 |
| τ²-Bench Telecom (AA run) | 73.4 | 99.1 |
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.