A focused vision-language model built around spatial reasoning and object localization within images. It takes text queries paired with images and returns textual descriptions of where things are — grounding language in visual space. At 3B parameters, it sits in a compact range that trades raw generality for targeted capability in detection-oriented tasks.