SMark Opens Source OXR for Advanced Document Parsing

York, NY (PRUnderground) October 20th, 2026, The team behind SoMark has open-sourced OXR, a groundbreaking 1.3-billion-parameter document parsing model. OXR is designed to transform PDFs and document images into agent-friendly data, making it a valuable tool for developers and researchers.

OXR stands out as the first document parser to incorporate a recursive self-improvement (RSI) loop during post-training. This innovative approach uses parsing errors and uncertain results to plan the next training round and generate targeted training data. According to the evaluation report, OXR excels on three key benchmarks: it ranks first on olmOCR-Bench, OmniDocBench, and DocParsingBench. Additionally, in end-to-end speed tests on a single RTX PRO 6000, OXR outperformed 21 other models, ranking first on the same 100-page OmniDocBench subset.

The model's capabilities extend beyond basic text recognition. OXR can identify 13 document element types, including tables, formulas, chemical equations, chemical structures, and code blocks. It is particularly adept at handling challenging documents and layouts, such as A3 exam papers, newspapers, vertical text, and handwritten content. OXR preserves the document structure and reading order in its output, with layout coordinates linking parsed content to its position on the original page. This feature supports downstream retrieval and agent workflows.

OXR is intended for AI builders developing LLM applications, RAG systems, agents, and document workflows, as well as document AI researchers and open-source developers. The model's task-level architectural redesign supports faster parsing by running recognition and reading order in parallel within the same model. Layout analysis remains independent, allowing a lighter model to handle that task while preserving layout accuracy.

The source code for OXR is available on GitHub, and its model weights are available on Hugging Face. The project uses an Apache 2.0-based custom license with additional terms, as specified in the project's LICENSE file. This open-source release marks a significant advancement in document parsing technology, providing developers with a powerful tool to enhance their applications and workflows.