Description
DeepSeek-OCR is a model that compresses long text by treating it as an image. This optical compression uses far fewer vision tokens to represent documents, unlocking new levels of efficiency for long-context tasks while delivering powerful OCR capabilities.
