| Benchmark | DeepSeek-R1 | Muse Spark |
|---|---|---|
| AA Agentic Index | 3.1 | 28.7 |
| AA Intelligence | 18.6 | 44.3 |
| AA-LCR | 56 | 77 |
| AA-Omniscience | -31.3 | 7.2 |
| ARC-AGI-2 | 1.3 | 42.5 |
| Artificial Analysis Coding Index | 24.6 | 58.6 |
| critpt | 0.6 | 11.3 |
| Fortress | 74.4 | 20.2 |
| gdpval | 1.5 | 32.3 |
| GPQA Diamond | 81 | 89.5 |
| HLE | 17.7 | 58.4 |
| IFBench | 39 | 75.9 |
| multichallenge | 45 | 75.5 |
| MultiNRC | 27.6 | 59 |
| OmniScience Accuracy | 30.7 | 49.6 |
| OmniScience Non-Hallucination | 10.5 | 26.8 |
| scicode | 35.7 | 51.5 |
| simpleqa_verified | 27.4 | 66.3 |
| SWE-bench Verified | 57.6 | 77.4 |
| TauBench V3 - Banking | 6.4 | 19.6 |
| Terminal-Bench 2.1 | 19.1 | 62.2 |
| Terminal-Bench Hard | 6.1 | 45.5 |
| τ²-Bench Telecom (AA run) | 11.4 | 91.5 |
| τ³-Bench | 6.4 | 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.