| Benchmark | Claude Sonnet 4.5 | GLM-5.2 Full Open Source |
|---|---|---|
| AA Agentic Index | 50.6 | 45.7 |
| AA Intelligence | 37.4 | 52.6 |
| AA-LCR | 68.3 | 76.7 |
| AA-Omniscience | -0.1 | 4.4 |
| arc_agi_1 | 63.7 | 77 |
| ARC-AGI-2 | 13.6 | 22.8 |
| arena_elo | 1456 | 1471 |
| arena_text_factuality | 1448 | 1460 |
| Artificial Analysis Coding Index | 52.1 | 68.8 |
| coding_arena_elo | 1392 | 1582 |
| critpt | 1.1 | 20.9 |
| gdpval | 40.4 | 50.3 |
| GPQA Diamond | 83.4 | 91.2 |
| HLE | 19.8 | 54.7 |
| HLE (with tools) | 32 | 54.7 |
| HMMT 2025 | 74.6 | 94.4 |
| hmmt_nov_2025 | 81.7 | 94.4 |
| IFBench | 57.3 | 73.3 |
| imo_answer_bench | 65.9 | 91 |
| MCP Atlas | 59.5 | 82.6 |
| OmniScience Accuracy | 32.9 | 24.3 |
| OmniScience Non-Hallucination | 50.9 | 73.7 |
| scicode | 45 | 50.5 |
| simplebench | 54.3 | 58.8 |
| simpleqa_verified | 23.6 | 38.1 |
| SWE-bench Pro | 43.6 | 62.1 |
| TauBench V3 - Banking | 24.5 | 34.6 |
| Terminal-Bench 2.1 | 55.8 | 82.7 |
| Terminal-Bench Hard | 35.6 | 50.8 |
| toolathlon | 41 | 48.2 |
| τ²-Bench Telecom (AA run) | 78.1 | 99.1 |
| τ³-Bench | 19 | 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.