| Benchmark | GLM-5.1 | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 66 | 39.7 |
| AA Intelligence | 41 | 53.2 |
| AA-LCR | 68 | 81.3 |
| AA-Omniscience | 0.8 | 28.1 |
| arena_elo | 1468 | 1491 |
| arena_text_factuality | 1458 | 1466 |
| Artificial Analysis Coding Index | 55.8 | 71.3 |
| coding_arena_elo | 1510 | 1539 |
| critpt | 4.6 | 15.1 |
| cybergym | 68.7 | 59 |
| Finance Agent v2 | 44.8 | 57.2 |
| gdpval | 49.5 | 43.7 |
| GPQA Diamond | 86.8 | 89.8 |
| HLE | 52.3 | 62.1 |
| MCP Atlas | 75.6 | 88.1 |
| OmniScience Accuracy | 25.2 | 52 |
| OmniScience Non-Hallucination | 70.1 | 50 |
| scicode | 43.8 | 58.2 |
| SWE-bench Pro | 58.4 | 61.5 |
| TauBench V3 - Banking | 13.6 | 31.8 |
| Terminal-Bench 2.1 | 63.5 | 80 |
| toolathlon | 40.7 | 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.