A nimble, conversational model that punches above its weight for an 8B parameter system. It handles everyday instruction-following, summarization, and light reasoning with solid reliability, though it can stumble on complex multi-step logic or nuanced tasks where larger models have a clear edge. Think of it as a capable generalist that runs lean — practical for deployment where compute matters.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| MMLU-Pro | 31.1 | accuracy | 26d ago |
| MuSR | 8.6 | accuracy | 26d ago |
| IFEval | 49.2 | accuracy | 26d ago |
| GPQA Diamond | 8.7 | accuracy | 26d ago |
| BBH | 29.4 | accuracy | 26d ago |
| MATH | 15.6 | accuracy | 26d ago |