| Benchmark | Muse Spark 1.1 | Qwen3.5 35B A3B |
|---|---|---|
| AA Agentic Index | 39.7 | 44.1 |
| AA Intelligence | 53.2 | 29.9 |
| AA-LCR | 81.3 | 68.3 |
| AA-Omniscience | 28.1 | -48.1 |
| Artificial Analysis Coding Index | 71.3 | 37 |
| baby_vision | 76.3 | 38.4 |
| coding_arena_elo | 1539 | 1250 |
| critpt | 15.1 | 0.9 |
| gdpval | 43.7 | 20.3 |
| GPQA Diamond | 89.8 | 84.5 |
| HLE | 62.1 | 47.4 |
| MCP Atlas | 88.1 | 62.4 |
| multichallenge | 75.3 | 60 |
| OmniScience Accuracy | 52 | 20.1 |
| OmniScience Non-Hallucination | 50 | 14.6 |
| OSWorld-Verified | 80.8 | 54.5 |
| scicode | 58.2 | 37.7 |
| SWE-bench Pro | 61.5 | 44.6 |
| TauBench V3 - Banking | 31.8 | 4.9 |
| Terminal-Bench 2.1 | 80 | 40.8 |
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.