| Benchmark | GPT-5.4 | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 58.2 | 39.7 |
| AA Intelligence | 53.1 | 53.2 |
| AA-LCR | 77.7 | 81.3 |
| AA-Omniscience | 5.8 | 28.1 |
| arena_elo | 1477 | 1491 |
| arena_text_factuality | 1476 | 1466 |
| Artificial Analysis Coding Index | 71.1 | 71.3 |
| baby_vision | 49.7 | 76.3 |
| coding_arena_elo | 1457 | 1539 |
| critpt | 23.4 | 15.1 |
| cybergym | 66.3 | 59 |
| deepsearchqa_f1 | 78.6 | 84.9 |
| DeepSWE 1.1 | 52 | 53.3 |
| gdpval | 50.1 | 43.7 |
| GPQA Diamond | 93 | 89.8 |
| HLE | 43.7 | 62.1 |
| livebench_agentic_coding | 53.8 | 58.5 |
| livebench_coding | 77.5 | 77.2 |
| livebench_data_analysis | 79.3 | 72.5 |
| livebench_instruction_following | 70.2 | 69.6 |
| livebench_language | 82.6 | 74.3 |
| livebench_math | 94.2 | 87.1 |
| livebench_reasoning | 88.1 | 87.7 |
| MCP Atlas | 70.6 | 88.1 |
| MultiNRC | 58.3 | 65.6 |
| OmniScience Accuracy | 50.9 | 52 |
| OmniScience Non-Hallucination | 17.4 | 50 |
| OSWorld-Verified | 75 | 80.8 |
| scicode | 56.6 | 58.2 |
| SWE-bench Pro | 59.1 | 61.5 |
| TauBench V3 - Banking | 39.6 | 31.8 |
| Terminal-Bench 2.1 | 78.3 | 80 |
| toolathlon | 54.6 | 75.6 |
| VisualToolBench | 29.2 | 44.8 |
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.