| Benchmark | GPT-5 | Muse Spark |
|---|---|---|
| AA Agentic Index | 49.7 | 28.7 |
| AA Intelligence | 35.3 | 44.3 |
| AA-LCR | 76.3 | 77 |
| AA-Omniscience | -8.7 | 7.2 |
| ARC-AGI-2 | 7.5 | 42.5 |
| arena_vision | 1212 | 1295 |
| Artificial Analysis Coding Index | 38.9 | 58.6 |
| coding_arena_elo | 1419 | 1540 |
| critpt | 5.7 | 11.3 |
| ERQA | 65.7 | 64.7 |
| Fortress | 17 | 20.2 |
| frontiermath_tier_4 | 12.5 | 14.6 |
| gdpval | 32.5 | 32.3 |
| GPQA Diamond | 87.3 | 89.5 |
| HLE | 28.5 | 58.4 |
| IFBench | 73.1 | 75.9 |
| MMMU-Pro | 78.4 | 80.5 |
| multichallenge | 71.1 | 75.5 |
| MultiNRC | 52.1 | 59 |
| OmniScience Accuracy | 40.3 | 49.6 |
| OmniScience Non-Hallucination | 21 | 26.8 |
| scicode | 43 | 51.5 |
| simpleqa_verified | 50.6 | 66.3 |
| SWE-bench Pro | 41.8 | 55 |
| SWE-bench Verified | 74.9 | 77.4 |
| TauBench V3 - Banking | 22.1 | 19.6 |
| Terminal-Bench 2.0 | 35.2 | 59 |
| Terminal-Bench 2.1 | 35.2 | 62.2 |
| Terminal-Bench Hard | 37.9 | 45.5 |
| τ²-Bench Telecom (AA run) | 86.5 | 91.5 |
| τ³-Bench | 19.6 | 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.