A compact, focused model built specifically for optical character recognition tasks. At 1B parameters, it stays lean and purposeful — processing images containing text and returning the extracted content. Its open-weight nature means it can run locally without API calls, which suits workflows where document throughput or data privacy matters.