| Benchmark | GLM-5.1 | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 66 | 27.5 |
| AA Intelligence | 41 | 37.7 |
| AA-LCR | 68 | 73.3 |
| AA-Omniscience | 0.8 | -20 |
| aime_2026 | 95.8 | 94.1 |
| Artificial Analysis Coding Index | 55.8 | 53.7 |
| critpt | 4.6 | 1.1 |
| gdpval | 49.5 | 32 |
| GPQA Diamond | 86.8 | 87.8 |
| HLE | 52.3 | 24 |
| HMMT 2025 | 94 | 90.7 |
| hmmt_feb_2026 | 89.4 | 84.3 |
| hmmt_nov_2025 | 94 | 90.7 |
| IFBench | 76.3 | 69.1 |
| imo_answer_bench | 83.8 | 80.8 |
| MMLU-Pro | 86 | 86.2 |
| nl2repo | 42.7 | 36.2 |
| OmniScience Accuracy | 25.2 | 19.6 |
| OmniScience Non-Hallucination | 70.1 | 50.7 |
| scicode | 43.8 | 39.8 |
| SWE-bench Multilingual | 73.3 | 71.3 |
| SWE-bench Pro | 58.4 | 53.5 |
| TauBench V3 - Banking | 13.6 | 16.7 |
| Terminal-Bench 2.0 | 69 | 59.3 |
| Terminal-Bench 2.1 | 63.5 | 60.7 |
| Terminal-Bench Hard | 43.2 | 34.8 |
| τ²-Bench Telecom (AA run) | 97.7 | 94.2 |
| τ³-Bench | 70.6 | 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.