MonkeyOCR: Document Parsing with a Structure-Recognition-Relation Triplet Paradigm

MonkeyOCR: Document Parsing with a Structure-Recognition-Relation Triplet Paradigm
Zhang Li, Yuliang Liu, Qiang Liu, Zhiyin Ma, Ziyang Zhang, Shuo Zhang, Zidun Guo, Jiarui Zhang, Xinyu Wang, Xiang Bai

Introduction

MonkeyOCR adopts a Structure-Recognition-Relation (SRR) triplet paradigm, which simplifies the multi-tool pipeline of modular approaches while avoiding the inefficiency of using large multimodal models for full-page document processing.

Compared with the pipeline-based method MinerU, our approach achieves an average improvement of 5.1% across nine types of Chinese and English documents, including a 15.0% gain on formulas and an 8.6% gain on tables.
Compared to end-to-end models, our 3B-parameter model achieves the best average performance on English documents, outperforming models such as Gemini 2.5 Pro and Qwen2.5 VL-72B.
For multi-page document parsing, our method reaches a processing speed of 0.84 pages per second, surpassing MinerU (0.65) and Qwen2.5 VL-7B (0.12).

MonkeyOCR currently does not support photographed documents, but we will continue to improve it in future updates. Stay tuned! Currently, our model is deployed on a single GPU, so if too many users upload files at the same time, issues like “This application is currently busy” may occur. We're actively working on supporting Ollama and other deployment solutions to ensure a smoother experience for more users. Additionally, please note that the processing time shown on the demo page does not reflect computation time alone—it also includes result uploading and other overhead. During periods of high traffic, this time may be longer. The inference speeds of MonkeyOCR, MinerU, and Qwen2.5 VL-7B were measured on an H800 GPU.

News

2025.06.12 🚀 The model’s trending on Hugging Face. Thanks for the love!
2025.06.05 🚀 We release MonkeyOCR, an English and Chinese documents parsing model.

Quick Start

Locally Install

1. Install MonkeyOCR

See the installation guide to set up your environment.

2. Download Model Weights

Download our model from Huggingface.

pip install huggingface_hub

python tools/download_model.py

You can also download our model from ModelScope.

pip install modelscope

python tools/download_model.py -t modelscope

3. Inference

You can parse a file or a directory containing PDFs or images using the following commands:

# Replace input_path with the path to a PDF or image or directory

# End-to-end parsing
python parse.py input_path

# Parse files in a dir with specific group page num
python parse.py input_path -g 20

# Single-task recognition (outputs markdown only)
python parse.py input_path -t text/formula/table

# Parse PDFs in input_path and split results by pages
python parse.py input_path -s

# Specify output directory and model config file
python parse.py input_path -o ./output -c config.yaml

More usage examples

# Single file processing
python parse.py input.pdf                           # Parse single PDF file
python parse.py input.pdf -o ./output               # Parse with custom output dir
python parse.py input.pdf -s                        # Parse PDF with page splitting
python parse.py image.jpg                           # Parse single image file

# Single task recognition
python parse.py image.jpg -t text                   # Text recognition from image
python parse.py image.jpg -t formula                # Formula recognition from image
python parse.py image.jpg -t table                  # Table recognition from image
python parse.py document.pdf -t text                # Text recognition from all PDF pages

# Folder processing (all files individually)
python parse.py /path/to/folder                     # Parse all files in folder
python parse.py /path/to/folder -s                  # Parse with page splitting
python parse.py /path/to/folder -t text             # Single task recognition for all files

# Multi-file grouping (batch processing by page count)
python parse.py /path/to/folder -g 5                # Group files with max 5 total pages
python parse.py /path/to/folder -g 10 -s            # Group files with page splitting
python parse.py /path/to/folder -g 8 -t text        # Group files for single task recognition

# Advanced configurations
python parse.py input.pdf -c model_configs.yaml     # Custom model configuration
python parse.py /path/to/folder -g 15 -s -o ./out   # Group files, split pages, custom output
python parse.py input.pdf --pred-abandon            # Enable predicting abandon elements

Tip

For Chinese scenarios, or cases where text, tables, etc. are mistakenly recognized as images, you can try using the following structure detection model: layout_zh.pt. (If the model is not found in model_weight/Structure/, you can download it manually.)

To use this model, update the configuration file model_configs.yaml as follows:

doclayout_yolo: Structure/layout_zh.pt

We have added support for the PP-DocLayout_plus-L, which offers improved performance over doclayout_yolo. Please refer to the Usage Guide.

To use this model, please update the configuration file model_configs.yaml as follows:

model: PP-DocLayout_plus-L

Output Results

MonkeyOCR generates three types of output files:

Processed Markdown File (your.md): The final parsed document content in markdown format, containing text, formulas, tables, and other structured elements.
Layout Results (your_layout.pdf): The layout results drawed on origin PDF.
Intermediate Block Results (your_middle.json): A JSON file containing detailed information about all detected blocks, including:
- Block coordinates and positions
- Block content and type information
- Relationship information between blocks

These files provide both the final formatted output and detailed intermediate results for further analysis or processing.

4. Gradio Demo

# Start demo
python demo/demo_gradio.py

5. Fast API

You can start the MonkeyOCR FastAPI service with the following command:

uvicorn api.main:app --port 8000

Once the API service is running, you can access the API documentation at http://localhost:8000/docs to explore available endpoints.

Tip

To improve API concurrency performance, consider configuring the inference backend as lmdeploy_queue or vllm_queue.

Docker Deployment

Navigate to the docker directory:
```
cd docker
```
Prerequisite: Ensure NVIDIA GPU support is available in Docker (via nvidia-docker2). If GPU support is not enabled, run the following to set up the environment:
```
bash env.sh
```
Build the Docker image:
```
docker compose build monkeyocr
```

Important

If your GPU is from the 30/40-series, V100, or similar, please build the patched Docker image for LMDeploy compatibility:

docker compose build monkeyocr-fix

Otherwise, you may encounter the following error: triton.runtime.errors.OutOfResources: out of resource: shared memory

Run the container with the Gradio demo (accessible on port 7860):
```
docker compose up monkeyocr-demo
```
Alternatively, start an interactive development environment:
```
docker compose run --rm monkeyocr-dev
```
Run the FastAPI service (accessible on port 7861):
```
docker compose up monkeyocr-api
```
Once the API service is running, you can access the API documentation at http://localhost:7861/docs to explore available endpoints.

Windows Support

See the Windows Support Guide for details.

Quantization

This model can be quantized using AWQ. Follow the instructions in the Quantization guide.

Benchmark Results

Here are the evaluation results of our model on OmniDocBench. MonkeyOCR-3B uses DocLayoutYOLO as the structure detection model, while MonkeyOCR-3B* uses our trained structure detection model with improved Chinese performance.

1. The end-to-end evaluation results of different tasks.

Model Type	Methods	Overall Edit↓		Text Edit↓		Formula Edit↓		Formula CDM↑		Table TEDS↑		Table Edit↓		Read Order Edit↓
Model Type	Methods	EN	ZH	EN	ZH	EN	ZH	EN	ZH	EN	ZH	EN	ZH	EN	ZH
Pipeline Tools	MinerU	0.150	0.357	0.061	0.215	0.278	0.577	57.3	42.9	78.6	62.1	0.180	0.344	0.079	0.292
	Marker	0.336	0.556	0.080	0.315	0.530	0.883	17.6	11.7	67.6	49.2	0.619	0.685	0.114	0.340
	Mathpix	0.191	0.365	0.105	0.384	0.306	0.454	62.7	62.1	77.0	67.1	0.243	0.320	0.108	0.304
	Docling	0.589	0.909	0.416	0.987	0.999	1	-	-	61.3	25.0	0.627	0.810	0.313	0.837
	Pix2Text	0.320	0.528	0.138	0.356	0.276	0.611	78.4	39.6	73.6	66.2	0.584	0.645	0.281	0.499
	Unstructured	0.586	0.716	0.198	0.481	0.999	1	-	-	0	0.06	1	0.998	0.145	0.387
	OpenParse	0.646	0.814	0.681	0.974	0.996	1	0.11	0	64.8	27.5	0.284	0.639	0.595	0.641
Expert VLMs	GOT-OCR	0.287	0.411	0.189	0.315	0.360	0.528	74.3	45.3	53.2	47.2	0.459	0.520	0.141	0.280
	Nougat	0.452	0.973	0.365	0.998	0.488	0.941	15.1	16.8	39.9	0	0.572	1.000	0.382	0.954
	Mistral OCR	0.268	0.439	0.072	0.325	0.318	0.495	64.6	45.9	75.8	63.6	0.600	0.650	0.083	0.284
	OLMOCR-sglang	0.326	0.469	0.097	0.293	0.455	0.655	74.3	43.2	68.1	61.3	0.608	0.652	0.145	0.277
	SmolDocling-256M	0.493	0.816	0.262	0.838	0.753	0.997	32.1	0.55	44.9	16.5	0.729	0.907	0.227	0.522
General VLMs	GPT4o	0.233	0.399	0.144	0.409	0.425	0.606	72.8	42.8	72.0	62.9	0.234	0.329	0.128	0.251
	Qwen2.5-VL-7B	0.312	0.406	0.157	0.228	0.351	0.574	79.0	50.2	76.4	72.2	0.588	0.619	0.149	0.203
	InternVL3-8B	0.314	0.383	0.134	0.218	0.417	0.563	78.3	49.3	66.1	73.1	0.586	0.564	0.118	0.186
Mix	MonkeyOCR-3B [Weight]	0.140	0.297	0.058	0.185	0.238	0.506	78.7	51.4	80.2	77.7	0.170	0.253	0.093	0.244
Mix	MonkeyOCR-3B* [Weight]	0.154	0.277	0.073	0.134	0.255	0.529	78.5	50.8	78.2	76.2	0.182	0.262	0.105	0.183

2. The end-to-end text recognition performance across 9 PDF page types.

Model Type	Models	Book	Slides	Financial Report	Textbook	Exam Paper	Magazine	Academic Papers	Notes	Newspaper	Overall
Pipeline Tools	MinerU	0.055	0.124	0.033	0.102	0.159	0.072	0.025	0.984	0.171	0.206
	Marker	0.074	0.340	0.089	0.319	0.452	0.153	0.059	0.651	0.192	0.274
	Mathpix	0.131	0.220	0.202	0.216	0.278	0.147	0.091	0.634	0.690	0.300
Expert VLMs	GOT-OCR	0.111	0.222	0.067	0.132	0.204	0.198	0.179	0.388	0.771	0.267
Expert VLMs	Nougat	0.734	0.958	1.000	0.820	0.930	0.830	0.214	0.991	0.871	0.806
General VLMs	GPT4o	0.157	0.163	0.348	0.187	0.281	0.173	0.146	0.607	0.751	0.316
	Qwen2.5-VL-7B	0.148	0.053	0.111	0.137	0.189	0.117	0.134	0.204	0.706	0.205
	InternVL3-8B	0.163	0.056	0.107	0.109	0.129	0.100	0.159	0.150	0.681	0.188
Mix	MonkeyOCR-3B [Weight]	0.046	0.120	0.024	0.100	0.129	0.086	0.024	0.643	0.131	0.155
Mix	MonkeyOCR-3B* [Weight]	0.054	0.203	0.038	0.112	0.138	0.111	0.032	0.194	0.136	0.120

3. Comparing MonkeyOCR with closed-source and extra large open-source VLMs.

Visualization Demo

Get a Quick Hands-On Experience with Our Demo: http://vlrlabmonkey.xyz:7685

Our demo is simple and easy to use:

Upload a PDF or image.

Click “Parse (解析)” to let the model perform structure detection, content recognition, and relationship prediction on the input document. The final output will be a markdown-formatted version of the document.

Select a prompt and click “Test by prompt” to let the model perform content recognition on the image based on the selected prompt.

Support diverse Chinese and English PDF types

Example for formula document

Example for table document

Example for newspaper

Example for financial report

Citing MonkeyOCR

If you wish to refer to the baseline results published here, please use the following BibTeX entries:

@misc{li2025monkeyocrdocumentparsingstructurerecognitionrelation,
      title={MonkeyOCR: Document Parsing with a Structure-Recognition-Relation Triplet Paradigm}, 
      author={Zhang Li and Yuliang Liu and Qiang Liu and Zhiyin Ma and Ziyang Zhang and Shuo Zhang and Zidun Guo and Jiarui Zhang and Xinyu Wang and Xiang Bai},
      year={2025},
      eprint={2506.05218},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2506.05218}, 
}

Acknowledgments

We would like to thank MinerU, DocLayout-YOLO, PyMuPDF, layoutreader, Qwen2.5-VL, LMDeploy, PP-StructureV3, PP-DocLayout_plus-L and InternVL3 for providing base code and models, as well as their contributions to this field. We also thank M6Doc, DocLayNet, CDLA, D4LA, DocGenome, PubTabNet, and UniMER-1M for providing valuable datasets. We also thank everyone who contributed to this open-source effort.

Alternative Models to Explore

If you find that our model doesn’t fully meet your needs, feel free to try out the following two recently released awesome models:

PP-StructureV3

MinerU 2.0

Copyright

Please don’t hesitate to share your valuable feedback — it’s a key motivation that drives us to continuously improve our framework. The current technical report only presents the results of the 3B model. Our model is intended for non-commercial use. If you are interested in larger one, please contact us at xbai@hust.edu.cn or ylliu@hust.edu.cn.

TruthHun/MonkeyOCR