A compact, specialized embedding model fine-tuned for geographic text. It operates within a tight 512-token window, so it works best with short passages rather than long documents. Its strength lies in producing dense vector representations of location-related content, making it useful for semantic similarity tasks in geo-focused applications.