| Benchmark | GLM-5.1 | Muse Spark |
|---|---|---|
| AA Agentic Index | 66 | 28.7 |
| AA Intelligence | 41 | 44.3 |
| AA-LCR | 68 | 77 |
| AA-Omniscience | 0.8 | 7.2 |
| arena_elo | 1468 | 1488 |
| arena_text_factuality | 1458 | 1467 |
| Artificial Analysis Coding Index | 55.8 | 58.6 |
| coding_arena_elo | 1510 | 1540 |
| critpt | 4.6 | 11.3 |
| cybergym | 68.7 | 43.5 |
| frontiermath_tier_4 | 12.5 | 14.6 |
| gdpval | 49.5 | 32.3 |
| GPQA Diamond | 86.8 | 89.5 |
| HLE | 52.3 | 58.4 |
| IFBench | 76.3 | 75.9 |
| itbenchSre | 40.3 | 16.2 |
| MCP Atlas | 75.6 | 82.2 |
| OmniScience Accuracy | 25.2 | 49.6 |
| OmniScience Non-Hallucination | 70.1 | 26.8 |
| scicode | 43.8 | 51.5 |
| simpleqa_verified | 38.1 | 66.3 |
| SWE-bench Pro | 58.4 | 55 |
| TauBench V3 - Banking | 13.6 | 19.6 |
| Terminal-Bench 2.0 | 69 | 59 |
| Terminal-Bench 2.1 | 63.5 | 62.2 |
| Terminal-Bench Hard | 43.2 | 45.5 |
| τ²-Bench Telecom (AA run) | 97.7 | 91.5 |
| τ³-Bench | 70.6 | 19.6 |
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.