| Benchmark | GPT-5.4 | Muse Spark |
|---|---|---|
| AA Agentic Index | 58.2 | 28.7 |
| AA Intelligence | 53.1 | 44.3 |
| AA-LCR | 77.7 | 77 |
| AA-Omniscience | 5.8 | 7.2 |
| ARC-AGI-2 | 74 | 42.5 |
| arena_elo | 1477 | 1488 |
| arena_text_factuality | 1476 | 1467 |
| arena_vision | 1277 | 1295 |
| Artificial Analysis Coding Index | 71.1 | 58.6 |
| coding_arena_elo | 1457 | 1540 |
| critpt | 23.4 | 11.3 |
| cybergym | 66.3 | 43.5 |
| deepsearchqa_f1 | 78.6 | 74.8 |
| frontiermath_tier_4 | 27.1 | 14.6 |
| gdpval | 50.1 | 32.3 |
| GPQA Diamond | 93 | 89.5 |
| HLE | 43.7 | 58.4 |
| IFBench | 73.9 | 75.9 |
| itbenchSre | 34.5 | 16.2 |
| MCP Atlas | 70.6 | 82.2 |
| MMMU-Pro | 81.2 | 80.5 |
| MultiNRC | 58.3 | 59 |
| OmniScience Accuracy | 50.9 | 49.6 |
| OmniScience Non-Hallucination | 17.4 | 26.8 |
| scicode | 56.6 | 51.5 |
| simpleqa_verified | 45.3 | 66.3 |
| SWE-bench Pro | 59.1 | 55 |
| TauBench V3 - Banking | 39.6 | 19.6 |
| Terminal-Bench 2.0 | 75.1 | 59 |
| Terminal-Bench 2.1 | 78.3 | 62.2 |
| Terminal-Bench Hard | 57.6 | 45.5 |
| τ²-Bench Telecom (AA run) | 87.1 | 91.5 |
| τ³-Bench | 72.9 | 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.