| Benchmark | Claude Opus 4.5 | Muse Spark |
|---|---|---|
| AA Agentic Index | 59.6 | 28.7 |
| AA Intelligence | 41.9 | 44.3 |
| AA-LCR | 76 | 77 |
| AA-Omniscience | 14 | 7.2 |
| ARC-AGI-2 | 37.6 | 42.5 |
| arena_elo | 1473 | 1488 |
| arena_text_factuality | 1458 | 1467 |
| Artificial Analysis Coding Index | 47.8 | 58.6 |
| coding_arena_elo | 1495 | 1540 |
| critpt | 4.6 | 11.3 |
| cybergym | 50.6 | 43.5 |
| deepsearchqa_f1 | 76.1 | 74.8 |
| ERQA | 46.8 | 64.7 |
| Fortress | 13.6 | 20.2 |
| frontiermath_tier_4 | 4.2 | 14.6 |
| gdpval | 47.3 | 32.3 |
| GPQA Diamond | 87 | 89.5 |
| HLE | 30.8 | 58.4 |
| IFBench | 58 | 75.9 |
| MCP Atlas | 69.8 | 82.2 |
| MMMU-Pro | 74 | 80.5 |
| multichallenge | 59 | 75.5 |
| MultiNRC | 48.6 | 59 |
| OmniScience Accuracy | 46.6 | 49.6 |
| OmniScience Non-Hallucination | 39 | 26.8 |
| scicode | 50 | 51.5 |
| simpleqa_verified | 41.8 | 66.3 |
| SimpleVQA | 69.7 | 0.7 |
| SWE-bench Pro | 57.1 | 55 |
| SWE-bench Verified | 81.5 | 77.4 |
| Terminal-Bench 2.0 | 59.3 | 59 |
| Terminal-Bench Hard | 47 | 45.5 |
| ZeroBench | 3 | 0.3 |
| τ²-Bench Telecom (AA run) | 89.5 | 91.5 |
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.