| Benchmark | Gemma 4 26B A4B | Muse Spark 1.1 |
|---|---|---|
| AA Agentic Index | 32.1 | 39.7 |
| AA Intelligence | 26.1 | 53.2 |
| AA-LCR | 61.7 | 81.3 |
| AA-Omniscience | -50.8 | 28.1 |
| arena_elo | 1438 | 1491 |
| arena_text_factuality | 1445 | 1466 |
| Artificial Analysis Coding Index | 39.3 | 71.3 |
| coding_arena_elo | 1361 | 1539 |
| critpt | 0 | 15.1 |
| gdpval | 25.7 | 43.7 |
| GPQA Diamond | 82.3 | 89.8 |
| HLE | 19.3 | 62.1 |
| MCP Atlas | 50 | 88.1 |
| OmniScience Accuracy | 19.1 | 52 |
| OmniScience Non-Hallucination | 13.6 | 50 |
| scicode | 40.3 | 58.2 |
| SWE-bench Pro | 13.8 | 61.5 |
| TauBench V3 - Banking | 12 | 31.8 |
| Terminal-Bench 2.1 | 39 | 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.