| Benchmark | Muse Spark | Qwen3.5 397B A17B |
|---|---|---|
| AA Agentic Index | 28.7 | 53.3 |
| AA Intelligence | 44.3 | 34.3 |
| AA-LCR | 77 | 72.7 |
| AA-Omniscience | 7.2 | -30.8 |
| arena_elo | 1488 | 1442 |
| arena_text_factuality | 1467 | 1445 |
| arena_vision | 1295 | 1249 |
| Artificial Analysis Coding Index | 58.6 | 48.2 |
| coding_arena_elo | 1540 | 1399 |
| critpt | 11.3 | 1.7 |
| ERQA | 64.7 | 67.5 |
| gdpval | 32.3 | 35.8 |
| GPQA Diamond | 89.5 | 89.3 |
| HLE | 58.4 | 29 |
| IFBench | 75.9 | 78.8 |
| itbenchSre | 16.2 | 34.1 |
| MMMU-Pro | 80.5 | 79 |
| multichallenge | 75.5 | 67.6 |
| OmniScience Accuracy | 49.6 | 30.8 |
| OmniScience Non-Hallucination | 26.8 | 17.3 |
| scicode | 51.5 | 42 |
| simpleqa_verified | 66.3 | 26 |
| SimpleVQA | 0.7 | 67.1 |
| SWE-bench Pro | 55 | 50.9 |
| SWE-bench Verified | 77.4 | 76.4 |
| TauBench V3 - Banking | 19.6 | 13.4 |
| Terminal-Bench 2.0 | 59 | 52.5 |
| Terminal-Bench 2.1 | 62.2 | 51.3 |
| Terminal-Bench Hard | 45.5 | 40.9 |
| τ²-Bench Telecom (AA run) | 91.5 | 95.6 |
| τ³-Bench | 19.6 | 13.4 |
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.