| Benchmark | GLM-5.2 Full Open Source | Qwen3.7 Max |
|---|---|---|
| AA Agentic Index | 45.7 | 30.9 |
| AA Intelligence | 52.6 | 46.7 |
| AA-LCR | 76.7 | 74.7 |
| AA-Omniscience | 4.4 | 13.5 |
| aime_2026 | 99.2 | 97 |
| arena_elo | 1471 | 1474 |
| arena_text_factuality | 1460 | 1479 |
| Artificial Analysis Coding Index | 68.8 | 66 |
| coding_arena_elo | 1582 | 1517 |
| critpt | 20.9 | 13.4 |
| DeepSWE | 46.2 | 18 |
| DeepSWE 1.1 | 44 | 21.6 |
| gdpval | 50.3 | 38.5 |
| GPQA Diamond | 91.2 | 92.4 |
| HLE | 54.7 | 41.4 |
| HLE (with tools) | 54.7 | 53.5 |
| hmmt_feb_2026 | 92.5 | 97.1 |
| hmmt_nov_2025 | 94.4 | 95 |
| IFBench | 73.3 | 80.5 |
| imo_answer_bench | 91 | 90 |
| itbenchSre | 42.7 | 42.5 |
| JobBench | 43.4 | 31.3 |
| livebench_agentic_coding | 51.8 | 43.6 |
| livebench_coding | 79.7 | 74.2 |
| livebench_data_analysis | 73.7 | 71.8 |
| livebench_instruction_following | 62.3 | 74 |
| livebench_language | 76.2 | 79.7 |
| livebench_math | 89.8 | 85.3 |
| livebench_reasoning | 78.6 | 83.3 |
| MCP Atlas | 82.6 | 76.4 |
| MLS Bench Lite | 40.4 | 31.7 |
| nl2repo | 48.9 | 47.2 |
| OmniScience Accuracy | 24.3 | 31.1 |
| OmniScience Non-Hallucination | 73.7 | 74.4 |
| scicode | 50.5 | 53.5 |
| simplebench | 58.8 | 70.4 |
| simpleqa_verified | 38.1 | 58.5 |
| SWE-bench Pro | 62.1 | 60.6 |
| TauBench V3 - Banking | 34.6 | 11.8 |
| Terminal-Bench 2.1 | 82.7 | 75 |
| Terminal-Bench Hard | 50.8 | 50.8 |
| τ²-Bench Telecom (AA run) | 99.1 | 94.7 |
| τ³-Bench | 26.8 | 10.9 |
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.