| Benchmark | Claude Opus 4.7 | GLM-5.2 Full Open Source |
|---|---|---|
| AA Agentic Index | 64.6 | 45.7 |
| AA Intelligence | 55 | 52.6 |
| AA-LCR | 75.3 | 76.7 |
| AA-Omniscience | 27.3 | 4.4 |
| aime_2026 | 95.8 | 99.2 |
| arc_agi_1 | 92 | 77 |
| ARC-AGI-2 | 75.8 | 22.8 |
| arena_elo | 1494 | 1471 |
| arena_text_factuality | 1474 | 1460 |
| Artificial Analysis Coding Index | 73.6 | 68.8 |
| coding_arena_elo | 1557 | 1582 |
| critpt | 12 | 20.9 |
| gdpval | 58.6 | 50.3 |
| GPQA Diamond | 94.2 | 91.2 |
| HiL-Bench | 41.7 | 43.7 |
| HLE | 54.7 | 54.7 |
| HLE (with tools) | 54.7 | 54.7 |
| hmmt_feb_2026 | 93.9 | 92.5 |
| IFBench | 58.6 | 73.3 |
| itbenchSre | 46.7 | 42.7 |
| livebench_agentic_coding | 50.7 | 51.8 |
| livebench_coding | 82.1 | 79.7 |
| livebench_data_analysis | 78.3 | 73.7 |
| livebench_instruction_following | 66.7 | 62.3 |
| livebench_language | 77.9 | 76.2 |
| livebench_math | 92.8 | 89.8 |
| livebench_reasoning | 87.2 | 78.6 |
| MCP Atlas | 79.1 | 82.6 |
| officeqa_pro | 80.6 | 41.4 |
| OmniScience Accuracy | 48.9 | 24.3 |
| OmniScience Non-Hallucination | 57.7 | 73.7 |
| scicode | 54.5 | 50.5 |
| simplebench | 61.7 | 58.8 |
| simpleqa_verified | 50.6 | 38.1 |
| SWE-bench Pro | 64.3 | 62.1 |
| TauBench V3 - Banking | 34.6 | 34.6 |
| Terminal-Bench 2.1 | 83.1 | 82.7 |
| Terminal-Bench Hard | 54.5 | 50.8 |
| toolathlon | 59.3 | 48.2 |
| τ²-Bench Telecom (AA run) | 88.6 | 99.1 |
| τ³-Bench | 28.9 | 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.