| Benchmark | GLM-5.1 | GLM-5.2 Full Open Source |
|---|---|---|
| AA Agentic Index | 66 | 45.7 |
| AA Intelligence | 41 | 52.6 |
| AA-LCR | 68 | 76.7 |
| AA-Omniscience | 0.8 | 4.4 |
| aime_2026 | 95.8 | 99.2 |
| arena_elo | 1468 | 1471 |
| arena_text_factuality | 1458 | 1460 |
| Artificial Analysis Coding Index | 55.8 | 68.8 |
| coding_arena_elo | 1510 | 1582 |
| critpt | 4.6 | 20.9 |
| DeepSWE | 18 | 46.2 |
| gdpval | 49.5 | 50.3 |
| GPQA Diamond | 86.8 | 91.2 |
| HLE | 52.3 | 54.7 |
| HLE (with tools) | 52.3 | 54.7 |
| HMMT 2025 | 94 | 94.4 |
| hmmt_feb_2026 | 89.4 | 92.5 |
| hmmt_nov_2025 | 94 | 94.4 |
| IFBench | 76.3 | 73.3 |
| imo_answer_bench | 83.8 | 91 |
| itbenchSre | 40.3 | 42.7 |
| MCP Atlas | 75.6 | 82.6 |
| nl2repo | 42.7 | 48.9 |
| OmniScience Accuracy | 25.2 | 24.3 |
| OmniScience Non-Hallucination | 70.1 | 73.7 |
| PostTrainBench | 20.1 | 34.3 |
| Program Bench | 50.9 | 63.7 |
| scicode | 43.8 | 50.5 |
| simplebench | 55.1 | 58.8 |
| simpleqa_verified | 38.1 | 38.1 |
| SWE-bench Pro | 58.4 | 62.1 |
| SWE-Marathon | 1 | 13 |
| TauBench V3 - Banking | 13.6 | 34.6 |
| Terminal-Bench 2.1 | 63.5 | 82.7 |
| Terminal-Bench Hard | 43.2 | 50.8 |
| tool_decathlon | 40.7 | 48.2 |
| toolathlon | 40.7 | 48.2 |
| τ²-Bench Telecom (AA run) | 97.7 | 99.1 |
| τ³-Bench | 70.6 | 26.8 |
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.