A remarkably compact model that punches above its weight class for a 1B parameter system. It handles surprisingly long contexts — up to 131K tokens — which is unusual at this scale. The trade-off is that, being a small supervised fine-tuned model, it will struggle with complex reasoning or nuanced tasks that larger models handle more gracefully.