| Benchmark | GLM-5.1 | GPT-5.5 |
|---|---|---|
| AA Agentic Index | 66 | 47.4 |
| AA Intelligence | 41 | 56.3 |
| AA-LCR | 68 | 79 |
| AA-Omniscience | 0.8 | 20.5 |
| aime_2026 | 95.8 | 100 |
| arena_elo | 1468 | 1482 |
| arena_text_factuality | 1458 | 1482 |
| Artificial Analysis Coding Index | 55.8 | 74.9 |
| browsecomp | 79.3 | 84.4 |
| coding_arena_elo | 1510 | 1457 |
| critpt | 4.6 | 27.1 |
| cybergym | 68.7 | 81.8 |
| DeepSWE | 18 | 70 |
| Finance Agent v2 | 44.8 | 51.8 |
| frontiermath_tier_4 | 12.5 | 35.4 |
| gdpval | 49.5 | 49.5 |
| GDPval-AA (Elo) | 1535 | 1769 |
| GPQA Diamond | 86.8 | 93.6 |
| HLE | 52.3 | 52.2 |
| HLE (with tools) | 52.3 | 52.2 |
| hmmt_feb_2026 | 89.4 | 98.5 |
| hmmt_nov_2025 | 94 | 96.5 |
| IFBench | 76.3 | 75.9 |
| itbenchSre | 40.3 | 45.8 |
| livebench | 70.2 | 80.7 |
| MCP Atlas | 75.6 | 82.8 |
| nl2repo | 42.7 | 50.7 |
| OmniScience Accuracy | 25.2 | 58 |
| OmniScience Non-Hallucination | 70.1 | 12 |
| PostTrainBench | 20.1 | 28.4 |
| Program Bench | 50.9 | 70.8 |
| scicode | 43.8 | 56.1 |
| simplebench | 55.1 | 69 |
| simpleqa_verified | 38.1 | 63.1 |
| SWE-bench Pro | 58.4 | 58.6 |
| SWE-Marathon | 1 | 14 |
| TauBench V3 - Banking | 13.6 | 39 |
| Terminal-Bench 2.0 | 69 | 82.7 |
| Terminal-Bench 2.1 | 63.5 | 84.3 |
| Terminal-Bench Hard | 43.2 | 60.6 |
| tool_decathlon | 40.7 | 55.6 |
| toolathlon | 40.7 | 55.6 |
| τ²-Bench Telecom (AA run) | 97.7 | 93.9 |
| τ³-Bench | 70.6 | 31.3 |
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.