Across 28 shared benchmarks, Muse Spark scores higher on 26 and Qwen3.5 122B A10B on 2. The widest gap is ZeroBench, where Muse Spark scores 0.3 against 0.1.
| Benchmark | Muse Spark | Qwen3.5 122B A10B |
|---|---|---|
| AA Agentic Index | 28.7 | 21.3 |
| AA Intelligence | 44.3 | 32.8 |
| AA-LCR | 77 | 70.3 |
| AA-Omniscience | 7.2 | -41.5 |
| arena_vision | 1306 | 1246 |
| Artificial Analysis Coding Index | 58.6 | 45.7 |
| coding_arena_elo | 1540 | 1358 |
| critpt | 11.3 | 0.9 |
| ERQA | 64.7 | 62 |
| gdpval | 32.3 | 24.3 |
| GPQA Diamond | 89.5 | 86.6 |
| HLE | 58.4 | 47.5 |
| IFBench | 75.9 | 76.1 |
| MMMU-Pro | 80.5 | 76.9 |
| multichallenge | 75.5 | 61.5 |
| OmniScience Accuracy | 49.6 | 24.4 |
| OmniScience Non-Hallucination | 26.8 | 12.9 |
| scicode | 51.5 | 42 |
| screenspot_pro_no_tools | 84.1 | 70.4 |
| SimpleVQA | 0.7 | 0.6 |
| SWE-bench Verified | 77.4 | 72 |
| TauBench V3 - Banking | 19.6 | 15.3 |
| Terminal-Bench 2.0 | 59 | 49.4 |
| Terminal-Bench 2.1 | 62.2 | 47.6 |
| Terminal-Bench Hard | 45.5 | 31.1 |
| ZeroBench | 0.3 | 0.1 |
| τ²-Bench Telecom (AA run) | 91.5 | 93.6 |
| τ³-Bench | 19.6 | 13.6 |
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.