RSI scores big! Domestic open-source tops global rankings, 1.3B OXR reshapes OCR

What happened: A domestic open-source document parsing model called OXR, with 1.3B parameters, topped three global benchmarks: OmniDocBench, olmOCR-Bench and DocParsingBench. Why it matters: The team says RSI training removes reliance on new manual annotation, and the layout analysis module is pluggable so businesses can swap in private models. A free API is offered without registration or keys. GitHub: https://github.com/SoMarkAI/OXR

A domestic open-source document parsing model called OXR, with 1.3B parameters, topped three global benchmarks: OmniDocBench, olmOCR-Bench and DocParsingBench.

The team says RSI training removes reliance on new manual annotation, and the layout analysis module is pluggable so businesses can swap in private models. A free API is offered without registration or keys. GitHub: https://github.com/SoMarkAI/OXR