| Benchmark | Llama 3.1 Instruct 405B | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 6.3 | 39.7 |
| AA Intelligence | 8.5 | 53.2 |
| AA-LCR | 24.7 | 81.3 |
| AA-Omniscience | -17.1 | 28.1 |
| Artificial Analysis Coding Index | 14.5 | 71.3 |
| critpt | 0 | 15.1 |
| Fortress | 20.6 | 12.4 |
| gdpval | 0 | 43.7 |
| GPQA Diamond | 51.5 | 89.8 |
| HLE | 4.2 | 62.1 |
| OmniScience Accuracy | 23.2 | 52 |
| OmniScience Non-Hallucination | 49 | 50 |
| scicode | 29.9 | 58.2 |
| SWE-bench Pro | 11.2 | 61.5 |
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.