| Benchmark | Claude Sonnet 4.5 | Muse Spark |
|---|---|---|
| AA Agentic Index | 50.6 | 28.7 |
| AA Intelligence | 37.4 | 44.3 |
| AA-LCR | 68.3 | 77 |
| AA-Omniscience | -0.1 | 7.2 |
| ARC-AGI-2 | 13.6 | 42.5 |
| arena_elo | 1456 | 1488 |
| arena_text_factuality | 1448 | 1467 |
| Artificial Analysis Coding Index | 52.1 | 58.6 |
| coding_arena_elo | 1392 | 1540 |
| critpt | 1.1 | 11.3 |
| Fortress | 17 | 20.2 |
| frontiermath_tier_4 | 4.2 | 14.6 |
| gdpval | 40.4 | 32.3 |
| GPQA Diamond | 83.4 | 89.5 |
| HLE | 19.8 | 58.4 |
| IFBench | 57.3 | 75.9 |
| MCP Atlas | 59.5 | 82.2 |
| MMMU-Pro | 68.7 | 80.5 |
| multichallenge | 55.3 | 75.5 |
| MultiNRC | 35.8 | 59 |
| OmniScience Accuracy | 32.9 | 49.6 |
| OmniScience Non-Hallucination | 50.9 | 26.8 |
| scicode | 45 | 51.5 |
| simpleqa_verified | 23.6 | 66.3 |
| SimpleVQA | 57.6 | 0.7 |
| SWE-bench Pro | 43.6 | 55 |
| SWE-bench Verified | 82 | 77.4 |
| TauBench V3 - Banking | 24.5 | 19.6 |
| Terminal-Bench 2.0 | 50 | 59 |
| Terminal-Bench 2.1 | 55.8 | 62.2 |
| Terminal-Bench Hard | 35.6 | 45.5 |
| τ²-Bench Telecom (AA run) | 78.1 | 91.5 |
| τ³-Bench | 19 | 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.