| Benchmark | GLM-5.1 | Qwen3.5 35B A3B |
|---|---|---|
| AA Agentic Index | 66 | 44.1 |
| AA Intelligence | 41 | 29.9 |
| AA-LCR | 68 | 68.3 |
| AA-Omniscience | 0.8 | -48.1 |
| AgentWorldBench - Android | 59.1 | 53.2 |
| AgentWorldBench - MCP | 67.6 | 57.9 |
| AgentWorldBench - OS | 59.1 | 56.3 |
| AgentWorldBench - Overall | 51.3 | 47.7 |
| AgentWorldBench - Search | 22.5 | 26 |
| AgentWorldBench - SWE | 52.1 | 47.6 |
| AgentWorldBench - Terminal | 47.3 | 46.1 |
| AgentWorldBench - Web | 51.5 | 47.1 |
| aime_2026 | 95.8 | 93.3 |
| Artificial Analysis Coding Index | 55.8 | 37 |
| browsecomp | 79.3 | 61 |
| coding_arena_elo | 1510 | 1250 |
| critpt | 4.6 | 0.9 |
| gdpval | 49.5 | 20.3 |
| GPQA Diamond | 86.8 | 84.5 |
| HLE | 52.3 | 47.4 |
| HMMT 2025 | 94 | 89.2 |
| hmmt_feb_2026 | 89.4 | 81.8 |
| hmmt_nov_2025 | 94 | 89.2 |
| IFBench | 76.3 | 72.5 |
| imo_answer_bench | 83.8 | 76.8 |
| itbenchSre | 40.3 | 21.5 |
| MCP Atlas | 75.6 | 62.4 |
| MMLU-Pro | 86 | 85.3 |
| nl2repo | 42.7 | 20.5 |
| OmniScience Accuracy | 25.2 | 20.1 |
| OmniScience Non-Hallucination | 70.1 | 14.6 |
| scicode | 43.8 | 37.7 |
| SWE-bench Multilingual | 73.3 | 60.3 |
| SWE-bench Pro | 58.4 | 44.6 |
| TauBench V3 - Banking | 13.6 | 4.9 |
| Terminal-Bench 2.0 | 69 | 40.5 |
| Terminal-Bench 2.1 | 63.5 | 40.8 |
| Terminal-Bench Hard | 43.2 | 26.5 |
| tool_decathlon | 40.7 | 28.7 |
| τ²-Bench Telecom (AA run) | 97.7 | 89.2 |
| τ³-Bench | 70.6 | 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.