| Benchmark | GLM-5.2 Full Open Source | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 45.7 | 27.5 |
| AA Intelligence | 52.6 | 37.7 |
| AA-LCR | 76.7 | 73.3 |
| AA-Omniscience | 4.4 | -20 |
| Agents' Last Exam | 23.8 | 27.3 |
| aime_2026 | 99.2 | 94.1 |
| Artificial Analysis Coding Index | 68.8 | 53.7 |
| critpt | 20.9 | 1.1 |
| DeepSWE 1.1 | 44 | 13.3 |
| gdpval | 50.3 | 32 |
| GPQA Diamond | 91.2 | 87.8 |
| HLE | 54.7 | 24 |
| HMMT 2025 | 94.4 | 90.7 |
| hmmt_feb_2026 | 92.5 | 84.3 |
| hmmt_nov_2025 | 94.4 | 90.7 |
| IFBench | 73.3 | 69.1 |
| imo_answer_bench | 91 | 80.8 |
| JobBench | 43.4 | 21.8 |
| livebench_agentic_coding | 51.8 | 39.3 |
| livebench_coding | 79.7 | 71.8 |
| livebench_data_analysis | 73.7 | 70.4 |
| livebench_instruction_following | 62.3 | 53.2 |
| livebench_language | 76.2 | 63.3 |
| livebench_math | 89.8 | 79.9 |
| livebench_reasoning | 78.6 | 70.3 |
| nl2repo | 48.9 | 36.2 |
| OmniScience Accuracy | 24.3 | 19.6 |
| OmniScience Non-Hallucination | 73.7 | 50.7 |
| scicode | 50.5 | 39.8 |
| SWE-bench Pro | 62.1 | 53.5 |
| TauBench V3 - Banking | 34.6 | 16.7 |
| Terminal-Bench 2.1 | 82.7 | 60.7 |
| Terminal-Bench Hard | 50.8 | 34.8 |
| τ²-Bench Telecom (AA run) | 99.1 | 94.2 |
| τ³-Bench | 26.8 | 15.3 |
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.