| Benchmark | GLM-5.2 Full Open Source | Qwen3.5 35B A3B |
|---|---|---|
| AA Agentic Index | 45.7 | 44.1 |
| AA Intelligence | 52.6 | 29.9 |
| AA-LCR | 76.7 | 68.3 |
| AA-Omniscience | 4.4 | -48.1 |
| aime_2026 | 99.2 | 93.3 |
| Artificial Analysis Coding Index | 68.8 | 37 |
| coding_arena_elo | 1582 | 1250 |
| critpt | 20.9 | 0.9 |
| gdpval | 50.3 | 20.3 |
| GPQA Diamond | 91.2 | 84.5 |
| HLE | 54.7 | 47.4 |
| HMMT 2025 | 94.4 | 89.2 |
| hmmt_feb_2026 | 92.5 | 81.8 |
| hmmt_nov_2025 | 94.4 | 89.2 |
| IFBench | 73.3 | 72.5 |
| imo_answer_bench | 91 | 76.8 |
| itbenchSre | 42.7 | 21.5 |
| MCP Atlas | 82.6 | 62.4 |
| nl2repo | 48.9 | 20.5 |
| OmniScience Accuracy | 24.3 | 20.1 |
| OmniScience Non-Hallucination | 73.7 | 14.6 |
| scicode | 50.5 | 37.7 |
| SWE-bench Pro | 62.1 | 44.6 |
| TauBench V3 - Banking | 34.6 | 4.9 |
| Terminal-Bench 2.1 | 82.7 | 40.8 |
| Terminal-Bench Hard | 50.8 | 26.5 |
| tool_decathlon | 48.2 | 28.7 |
| τ²-Bench Telecom (AA run) | 99.1 | 89.2 |
| τ³-Bench | 26.8 | 68.9 |
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.