SoMark Team Open-Sources 1.3B Document Parsing Model OXR, Topping Three Benchmarks
The SoMark team has open-sourced the document parsing model OXR. The main model has only 1.3B parameters and topped all three benchmark tests: OmniDocBench, olmOCR-Bench, and DocParsingBench. Its full parsing speed exceeds that of PaddleOCR-VL-1.6 and MinerU2.5-Pro, which were also tested in the same round. The model covers 13 types of document elements, including text, tables, formulas, chemical structure formulas, and code blocks. It can parse reading order and output layout coordinates, converting PDFs and document images into Markdown / JSON for downstream processes such as Agents and RAG.
Why read this
AI score 60
The original text provides the three benchmark top rankings and speed comparisons, allowing readers to judge where lightweight document parsing models fit in the RAG pipeline.
Topics
Body · AI translation
The document parsing track has welcomed another strong new contender.
OXR is now officially open source: the main model has only 1.3B parameters and tops all three benchmarks—OmniDocBench, olmOCR-Bench, and DocParsingBench—while its full parsing speed also surpasses that of PaddleOCR-VL-1.6 and MinerU2.5-Pro, which were tested alongside it, pushing a lightweight model to the top in both accuracy and efficiency at the same time. The SoMark team behind it comes from a domestic startup that has long been deeply engaged in the document intelligence field; having previously continued to refine parsing capabilities around its core product SoMark, it has now released the open-source model OXR.
Multi-column papers, newspapers, and exam papers also test reading order. Even if every region is recognized correctly, reading across the left and right columns or mixing body text with marginal notes will still distort the original meaning. OXR determines reading order while recognizing content and provides layout coordinates, preserving the position of content on the page.
Together, these capabilities constitute OXR's "Everything": covering 13 categories of document elements such as text, tables, formulas, chemical structure formulas, and code blocks, parsing PDFs and document images into Markdown / JSON for downstream processes such as Agents and RAG.
This article is provided by the SoMark team; QbitAI is authorized to repost it, and the views belong to the original author.