MOCR - A multimodal document parsing model jointly launched by Xiaohongshu and Huazhong University of Science and Technology.

What is MOCR?

MOCR (Multimodal OCR) is a multimodal document parsing model jointly developed by Huazhong University of Science and Technology and Xiaohongshu hi lab. It achieves breakthrough performance in document parsing and image reconstruction with only 3 parameters. Breaking away from the limitations of traditional OCR that only recognizes text, the model parses visual elements such as charts, formulas, and flowcharts into editable SVG code, realizing a new paradigm of "parsing everything." It ranks first among open-source models, and its image reconstruction capabilities surpass those of Gemini 3 Pro, bringing a paradigm shift to the field of document AI.

Main functions of MOCR

• Document full element analysisIt supports recognizing all page elements such as text, tables, formulas, and charts, outputting structured data while maintaining the reading order.
• Graphic to SVG code: Reconstruct visual content such as statistical charts, scientific illustrations, and UI layouts into editable SVG code.
• Multi-format input supportSupports parsing various document types, including PDFs, webpage screenshots, scanned documents, and photos taken with mobile phones.
• General visual capabilitiesThe model possesses general multimodal understanding capabilities, including visual question answering, visual localization, and image description.
• Dual-version modelIt provides a balanced version of dots.mocr and an SVG-optimized version of dots.mocr-svg to meet the needs of different scenarios.

Key information and usage requirements of MOCR

• Development TeamHuazhong University of Science and Technology × Xiaohongshu hi lab
• Model parameters3B (1.2B visual encoder + 1.5B language decoder)
• Model versiondots.mocr (balanced version), dots.mocr-svg (SVG enhanced version)
• Core InnovationParsing graphics into SVG code enables a new paradigm of "parse everything".
• PerformanceFirst in open-source document parsing, surpasses Gemini 3 Pro in image reconstruction.
• GPU: NVIDIA graphics card that supports CUDA (recommended for inference acceleration)
• Memory: Adjusted according to input resolution; high-resolution documents require more video memory.

MOCR's core advantages

• Small parameters, big capabilitiesWith only 3B parameters, it outperforms many large models, ranks first in open-source document parsing, and surpasses Gemini 3 Pro in image reconstruction.
• Full element analysisThe model breaks through the limitations of traditional OCR, which only recognizes text, and unifies the parsing of visual elements such as charts, formulas, and flowcharts into structured code.
• Graphic EditabilityConvert graphics to SVG code for lossless reconstruction and secondary editing, rather than simply cropping them into pixel images.
• Data Engine InnovationIt supports building multi-source data pipelines such as PDFs, web pages, and SVG assets, solving the problem of scarce graphical supervision signals.
• Evaluation Methodology InnovationThe model proposes an OCR Arena framework, using a strong VLM as the referee for reliable comparative evaluation.

How to use MOCR

• Environmental preparationCreate a Python 3.12 virtual environment, clone the GitHub repository, and install dependencies.
• Download ModelRun the download script to obtain the model weights. Note that the save path should not contain a period (.).
• Start service: Use vLLM to deploy model services and support GPU-accelerated inference.
• Document parsing: Calls the parsing script to process images or PDF files and outputs structured results.
• Graphics ConversionUse a dedicated SVG script to convert the chart into editable SVG code.
• Get Results: Get the generated JSON file containing bounding boxes, Markdown text, and a visual annotation diagram.

MOCR project address

• GitHub repositoryhttps://github.com/rednote-hilab/dots.mocr
• arXiv technical paper: https://arxiv.org/pdf/2603.13032
• Experience the demo online: https://dotsocr.xiaohongshu.com/

MOCR's Competitive Product Comparison

Application scenarios of MOCR

• Academic researchThis tool parses PDF papers, extracts formulas, and reconstructs charts, converting scanned papers into editable LaTeX format to reproduce scientific research charts and data.
• Finance and EconomicsAnalyze financial statements, extract data reports, and digitize charts, converting bar charts and line charts from PDF financial statements into usable data in Excel.
• Legal and political affairsReview contracts, digitize dossiers, identify documents, and extract key information from multi-page contracts in a structured manner while maintaining the integrity of the format.
• Educational PublishingDigital textbooks, test question banks, and blackboard recognition technologies convert complex formulas and charts in printed textbooks into electronic resources.
• HealthcareIt can analyze medical records, extract test reports, annotate medical images, and process documents containing chemical formulas and medical diagrams.