| Benchmark | GLM-5.2 Full Open Source | Qwen3.5 27B |
|---|---|---|
| AA Agentic Index | 45.7 | 54.6 |
| AA Intelligence | 52.6 | 34.6 |
| AA-LCR | 76.7 | 72.3 |
| AA-Omniscience | 4.4 | -44 |
| aime_2026 | 99.2 | 92.6 |
| Artificial Analysis Coding Index | 68.8 | 34.9 |
| coding_arena_elo | 1582 | 1357 |
| critpt | 20.9 | 0.9 |
| gdpval | 50.3 | 33 |
| GPQA Diamond | 91.2 | 85.8 |
| HLE | 54.7 | 48.5 |
| HMMT 2025 | 94.4 | 89.8 |
| hmmt_feb_2026 | 92.5 | 84.3 |
| hmmt_nov_2025 | 94.4 | 89.8 |
| IFBench | 73.3 | 76.5 |
| imo_answer_bench | 91 | 79.9 |
| itbenchSre | 42.7 | 35.5 |
| MCP Atlas | 82.6 | 68.4 |
| nl2repo | 48.9 | 27.3 |
| OmniScience Accuracy | 24.3 | 20.7 |
| OmniScience Non-Hallucination | 73.7 | 24.9 |
| scicode | 50.5 | 39.5 |
| SWE-bench Pro | 62.1 | 51.2 |
| Terminal-Bench Hard | 50.8 | 32.6 |
| tool_decathlon | 48.2 | 31.5 |
| τ²-Bench Telecom (AA run) | 99.1 | 93.9 |
| τ³-Bench | 26.8 | 68.4 |
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.