| Benchmark | Gemma 4 31B | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 14.4 | 39.7 |
| AA Intelligence | 29.7 | 53.2 |
| AA-LCR | 68.3 | 81.3 |
| AA-Omniscience | -47.9 | 28.1 |
| arena_elo | 1451 | 1491 |
| arena_text_factuality | 1449 | 1466 |
| Artificial Analysis Coding Index | 43.4 | 71.3 |
| coding_arena_elo | 1364 | 1539 |
| critpt | 1.4 | 15.1 |
| gdpval | 15.5 | 43.7 |
| GPQA Diamond | 85.7 | 89.8 |
| HLE | 26.5 | 62.1 |
| MCP Atlas | 57.2 | 88.1 |
| OmniScience Accuracy | 20 | 52 |
| OmniScience Non-Hallucination | 18.1 | 50 |
| scicode | 43.4 | 58.2 |
| SWE-bench Pro | 35.7 | 61.5 |
| TauBench V3 - Banking | 14.8 | 31.8 |
| Terminal-Bench 2.1 | 43.4 | 80 |
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.