| Benchmark | Claude Sonnet 4 | Muse Spark |
|---|---|---|
| AA Agentic Index | 39.2 | 28.7 |
| AA Intelligence | 29.8 | 44.3 |
| AA-LCR | 69.7 | 77 |
| AA-Omniscience | 0.2 | 7.2 |
| ARC-AGI-2 | 13.6 | 42.5 |
| arena_vision | 1207 | 1295 |
| Artificial Analysis Coding Index | 37.6 | 58.6 |
| critpt | 1.1 | 11.3 |
| Fortress | 24.4 | 20.2 |
| frontiermath_tier_4 | 0 | 14.6 |
| gdpval | 31.2 | 32.3 |
| GPQA Diamond | 78 | 89.5 |
| HLE | 10.7 | 58.4 |
| IFBench | 55 | 75.9 |
| MMMU-Pro | 62.4 | 80.5 |
| multichallenge | 57.1 | 75.5 |
| MultiNRC | 18.4 | 59 |
| OmniScience Accuracy | 22.7 | 49.6 |
| OmniScience Non-Hallucination | 70.9 | 26.8 |
| scicode | 40 | 51.5 |
| SWE-bench Pro | 42.7 | 55 |
| SWE-bench Verified | 80.2 | 77.4 |
| TauBench V3 - Banking | 16.7 | 19.6 |
| Terminal-Bench 2.1 | 36.3 | 62.2 |
| Terminal-Bench Hard | 31.1 | 45.5 |
| τ²-Bench Telecom (AA run) | 64.6 | 91.5 |
| τ³-Bench | 13.8 | 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.