| Benchmark | GLM-5.2 Full Open Source | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 45.7 | 20.8 |
| AA Intelligence | 52.6 | 39.4 |
| AA-LCR | 76.7 | 69 |
| AA-Omniscience | 4.4 | 2.4 |
| Agents' Last Exam | 23.8 | 33.6 |
| apexAgents | 33.7 | 22.4 |
| arena_elo | 1471 | 1458 |
| arena_text_factuality | 1460 | 1454 |
| Artificial Analysis Coding Index | 68.8 | 55.9 |
| critpt | 20.9 | 9.1 |
| DeepSWE 1.1 | 44 | 14.2 |
| gdpval | 50.3 | 22.2 |
| GPQA Diamond | 91.2 | 90.3 |
| HLE | 54.7 | 35.6 |
| hmmt_feb_2026 | 92.5 | 92.9 |
| IFBench | 73.3 | 79.1 |
| imo_answer_bench | 91 | 86 |
| JobBench | 43.4 | 27.6 |
| MCP Atlas | 82.6 | 73.2 |
| nl2repo | 48.9 | 41.1 |
| OmniScience Accuracy | 24.3 | 22.5 |
| OmniScience Non-Hallucination | 73.7 | 74.5 |
| scicode | 50.5 | 51.3 |
| SWE-bench Pro | 62.1 | 57.6 |
| TauBench V3 - Banking | 34.6 | 17.9 |
| Terminal-Bench 2.1 | 82.7 | 61 |
| Terminal-Bench Hard | 50.8 | 47 |
| τ²-Bench Telecom (AA run) | 99.1 | 93 |
| τ³-Bench | 26.8 | 17.9 |
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.