Florence 2 base is Microsoft's compact vision-language model that punches above its weight on image understanding tasks. It handles a broad range of visual tasks — captioning, object detection, OCR, and visual grounding — through a unified prompt-based interface rather than task-specific heads. Being the 'base' variant, it trades raw capability for efficiency, making it approachable for fine-tuning on custom vision tasks.