| Benchmark | GPT-5.2 | Muse Spark |
|---|---|---|
| AA Agentic Index | 60.2 | 28.7 |
| AA Intelligence | 43.3 | 44.3 |
| AA-LCR | 79.3 | 77 |
| AA-Omniscience | 0.3 | 7.2 |
| ARC-AGI-2 | 52.9 | 42.5 |
| arena_elo | 1437 | 1488 |
| arena_text_factuality | 1442 | 1467 |
| arena_vision | 1245 | 1295 |
| Artificial Analysis Coding Index | 48.7 | 58.6 |
| coding_arena_elo | 1418 | 1540 |
| critpt | 11.6 | 11.3 |
| deepsearchqa_f1 | 71.3 | 74.8 |
| Fortress | 17.5 | 20.2 |
| frontiermath_tier_4 | 18.8 | 14.6 |
| gdpval | 48.3 | 32.3 |
| GPQA Diamond | 92.4 | 89.5 |
| HLE | 45.5 | 58.4 |
| IFBench | 75.4 | 75.9 |
| MCP Atlas | 68 | 82.2 |
| MMMU-Pro | 79.5 | 80.5 |
| MultiNRC | 42.2 | 59 |
| OmniScience Accuracy | 44.3 | 49.6 |
| OmniScience Non-Hallucination | 38.4 | 26.8 |
| scicode | 52.1 | 51.5 |
| screenspot_pro_no_tools | 86.3 | 84.1 |
| simpleqa_verified | 38.9 | 66.3 |
| SimpleVQA | 55.8 | 0.7 |
| SWE-bench Pro | 55.6 | 55 |
| SWE-bench Verified | 80 | 77.4 |
| TauBench V3 - Banking | 11.1 | 19.6 |
| Terminal-Bench 2.0 | 54 | 59 |
| Terminal-Bench Hard | 62.2 | 45.5 |
| ZeroBench | 9 | 0.3 |
| τ²-Bench Telecom (AA run) | 98.7 | 91.5 |
| τ³-Bench | 12 | 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.