| Benchmark | GLM-5.2 Full Open Source | Muse Spark |
|---|---|---|
| AA Agentic Index | 45.7 | 28.7 |
| AA Intelligence | 52.6 | 44.3 |
| AA-LCR | 76.7 | 77 |
| AA-Omniscience | 4.4 | 7.2 |
| ARC-AGI-2 | 22.8 | 42.5 |
| arena_elo | 1471 | 1488 |
| arena_text_factuality | 1460 | 1467 |
| Artificial Analysis Coding Index | 68.8 | 58.6 |
| coding_arena_elo | 1582 | 1540 |
| critpt | 20.9 | 11.3 |
| gdpval | 50.3 | 32.3 |
| GPQA Diamond | 91.2 | 89.5 |
| HLE | 54.7 | 58.4 |
| IFBench | 73.3 | 75.9 |
| itbenchSre | 42.7 | 16.2 |
| MCP Atlas | 82.6 | 82.2 |
| OmniScience Accuracy | 24.3 | 49.6 |
| OmniScience Non-Hallucination | 73.7 | 26.8 |
| scicode | 50.5 | 51.5 |
| simpleqa_verified | 38.1 | 66.3 |
| SWE-bench Pro | 62.1 | 55 |
| TauBench V3 - Banking | 34.6 | 19.6 |
| Terminal-Bench 2.1 | 82.7 | 62.2 |
| Terminal-Bench Hard | 50.8 | 45.5 |
| τ²-Bench Telecom (AA run) | 99.1 | 91.5 |
| τ³-Bench | 26.8 | 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.