1 post tagged ocr.
A compact 0.9B multimodal model that handles document parsing, tables, and structured extraction without burning through compute resources.