| Benchmark | Claude Opus 4.6 | GLM-5.2 Full Open Source |
|---|---|---|
| AA Agentic Index | 67.6 | 45.7 |
| AA Intelligence | 44.9 | 52.6 |
| AA-LCR | 74.3 | 76.7 |
| AA-Omniscience | 13.7 | 4.4 |
| aime_2026 | 96.7 | 99.2 |
| apexAgents | 33 | 33.7 |
| arc_agi_1 | 93 | 77 |
| ARC-AGI-2 | 68.8 | 22.8 |
| arena_elo | 1497 | 1471 |
| arena_text_factuality | 1487 | 1460 |
| Artificial Analysis Coding Index | 48.1 | 68.8 |
| coding_arena_elo | 1545 | 1582 |
| critpt | 12.6 | 20.9 |
| gdpval | 55.9 | 50.3 |
| GPQA Diamond | 91.3 | 91.2 |
| HiL-Bench | 38.3 | 43.7 |
| HLE | 53.1 | 54.7 |
| HLE (with tools) | 62.7 | 54.7 |
| hmmt_feb_2026 | 96.2 | 92.5 |
| hmmt_nov_2025 | 96.3 | 94.4 |
| IFBench | 62.5 | 73.3 |
| imo_answer_bench | 75.3 | 91 |
| livebench_agentic_coding | 49 | 51.8 |
| livebench_coding | 78.2 | 79.7 |
| livebench_data_analysis | 69.9 | 73.7 |
| livebench_instruction_following | 63.3 | 62.3 |
| livebench_language | 83.3 | 76.2 |
| livebench_math | 89.3 | 89.8 |
| livebench_reasoning | 88.7 | 78.6 |
| MCP Atlas | 76.8 | 82.6 |
| nl2repo | 49.8 | 48.9 |
| officeqa_pro | 57.1 | 41.4 |
| OmniScience Accuracy | 47 | 24.3 |
| OmniScience Non-Hallucination | 37.2 | 73.7 |
| scicode | 52 | 50.5 |
| simplebench | 67.6 | 58.8 |
| simpleqa_verified | 46.5 | 38.1 |
| SWE-bench Pro | 57.3 | 62.1 |
| Terminal-Bench Hard | 65.4 | 50.8 |
| tool_decathlon | 47.2 | 48.2 |
| toolathlon | 56.8 | 48.2 |
| τ²-Bench Telecom (AA run) | 99.3 | 99.1 |
| τ³-Bench | 72.4 | 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.