| Benchmark | Muse Spark 1.1 | Qwen3.5 397B A17B |
|---|---|---|
| AA Agentic Index | 39.7 | 53.3 |
| AA Intelligence | 53.2 | 34.3 |
| AA-LCR | 81.3 | 72.7 |
| AA-Omniscience | 28.1 | -30.8 |
| arena_elo | 1491 | 1442 |
| arena_text_factuality | 1466 | 1445 |
| Artificial Analysis Coding Index | 71.3 | 48.2 |
| coding_arena_elo | 1539 | 1399 |
| critpt | 15.1 | 1.7 |
| gdpval | 43.7 | 35.8 |
| GPQA Diamond | 89.8 | 89.3 |
| HLE | 62.1 | 29 |
| multichallenge | 75.3 | 67.6 |
| OmniScience Accuracy | 52 | 30.8 |
| OmniScience Non-Hallucination | 50 | 17.3 |
| scicode | 58.2 | 42 |
| SWE-bench Pro | 61.5 | 50.9 |
| TauBench V3 - Banking | 31.8 | 13.4 |
| Terminal-Bench 2.1 | 80 | 51.3 |
| toolathlon | 75.6 | 38.3 |
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.