| Benchmark | GLM-5.2 Full Open Source | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 45.7 | 39.7 |
| AA Intelligence | 52.6 | 53.2 |
| AA-LCR | 76.7 | 81.3 |
| AA-Omniscience | 4.4 | 28.1 |
| arena_elo | 1471 | 1491 |
| arena_text_factuality | 1460 | 1466 |
| Artificial Analysis Coding Index | 68.8 | 71.3 |
| coding_arena_elo | 1582 | 1539 |
| critpt | 20.9 | 15.1 |
| DeepSWE 1.1 | 44 | 53.3 |
| gdpval | 50.3 | 43.7 |
| GPQA Diamond | 91.2 | 89.8 |
| HLE | 54.7 | 62.1 |
| JobBench | 43.4 | 54.7 |
| livebench_agentic_coding | 51.8 | 58.5 |
| livebench_coding | 79.7 | 77.2 |
| livebench_data_analysis | 73.7 | 72.5 |
| livebench_instruction_following | 62.3 | 69.6 |
| livebench_language | 76.2 | 74.3 |
| livebench_math | 89.8 | 87.1 |
| livebench_reasoning | 78.6 | 87.7 |
| MCP Atlas | 82.6 | 88.1 |
| OmniScience Accuracy | 24.3 | 52 |
| OmniScience Non-Hallucination | 73.7 | 50 |
| scicode | 50.5 | 58.2 |
| SWE-bench Pro | 62.1 | 61.5 |
| TauBench V3 - Banking | 34.6 | 31.8 |
| Terminal-Bench 2.1 | 82.7 | 80 |
| toolathlon | 48.2 | 75.6 |
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.