| Benchmark | DeepSeek-V3 | Muse Spark |
|---|---|---|
| AA Agentic Index | 1.6 | 28.7 |
| AA Intelligence | 15.4 | 44.3 |
| AA-LCR | 41.3 | 77 |
| AA-Omniscience | -37.6 | 7.2 |
| Artificial Analysis Coding Index | 23 | 58.6 |
| critpt | 0 | 11.3 |
| gdpval | 0 | 32.3 |
| GPQA Diamond | 68.4 | 89.5 |
| HLE | 5.2 | 58.4 |
| IFBench | 41 | 75.9 |
| multichallenge | 31.4 | 75.5 |
| OmniScience Accuracy | 25.4 | 49.6 |
| OmniScience Non-Hallucination | 23.3 | 26.8 |
| scicode | 35.8 | 51.5 |
| SWE-bench Verified | 42 | 77.4 |
| TauBench V3 - Banking | 4.7 | 19.6 |
| Terminal-Bench 2.1 | 16.9 | 62.2 |
| Terminal-Bench Hard | 15.2 | 45.5 |
| τ²-Bench Telecom (AA run) | 47.1 | 91.5 |
| τ³-Bench | 4.7 | 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.