Gemma 2 27B IT is Google's mid-sized open-weight model that punches above its parameter count, producing careful and coherent reasoning across a wide range of text tasks. It tends toward measured, well-structured responses and handles nuanced instruction-following reliably. The trade-off is that it requires meaningful compute to run locally, sitting in a tier where self-hosting demands real hardware.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| MuSR | 9.1 | accuracy | 24d ago |
| MATH | 23.9 | accuracy | 24d ago |
| GPQA Diamond | 16.7 | accuracy | 24d ago |
| BBH | 49.3 | accuracy | 24d ago |
| IFEval | 79.8 | accuracy | 24d ago |
| MMLU-Pro | 38.3 | accuracy | 24d ago |