| Benchmark | GLM-5.1 | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 66 | 57.8 |
| AA Intelligence | 41 | 61 |
| AA-LCR | 68 | 77.7 |
| AA-Omniscience | 0.8 | 22 |
| aime_2026 | 95.8 | 99.9 |
| arena_elo | 1468 | 1482 |
| arena_text_factuality | 1458 | 1474 |
| Artificial Analysis Coding Index | 55.8 | 78.3 |
| browsecomp | 79.3 | 90.4 |
| coding_arena_elo | 1510 | 1619 |
| critpt | 4.6 | 32.3 |
| cybergym | 68.7 | 83.6 |
| DeepSWE | 18 | 73 |
| Finance Agent v2 | 44.8 | 53.8 |
| gdpval | 49.5 | 61.1 |
| GPQA Diamond | 86.8 | 94.6 |
| HLE | 52.3 | 49.5 |
| HLE (with tools) | 52.3 | 58 |
| IFBench | 76.3 | 72.7 |
| itbenchSre | 40.3 | 56.2 |
| MCP Atlas | 75.6 | 83.6 |
| OmniScience Accuracy | 25.2 | 59.4 |
| OmniScience Non-Hallucination | 70.1 | 10.6 |
| PostTrainBench | 20.1 | 34.6 |
| Program Bench | 50.9 | 77.6 |
| scicode | 43.8 | 56.9 |
| simplebench | 55.1 | 64.8 |
| simpleqa_verified | 38.1 | 71.6 |
| SWE-bench Pro | 58.4 | 64.6 |
| SWE-Marathon | 1 | 39 |
| TauBench V3 - Banking | 13.6 | 44.3 |
| Terminal-Bench 2.1 | 63.5 | 89.5 |
| Terminal-Bench Hard | 43.2 | 65.9 |
| toolathlon | 40.7 | 58 |
| τ²-Bench Telecom (AA run) | 97.7 | 85.1 |
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.