MinerU2.5-Pro

MinerU2.5-Pro

Reads document page images and returns text, as a 1.2B-parameter model built for document parsing in Chinese and English. It suits teams that turn scanned pages into usable text.

Made by opendatalab on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: MinerU2.5-Pro is an image-to-text model from OpenDataLab with 1.2B parameters. Its card describes it as a document parsing model, built around a large-scale, data-focused approach. You give it a picture of a page, and it returns text. The record lists Chinese and English.

Who it suits: It suits developers and data teams who need text out of scanned or photographed documents. The card text provided here is short, so the details below are hedged.

Industry | Who uses it | How they use it | Example
Document processing | Data and operations teams | Turn scanned pages into text for later use | Reading a batch of scanned reports
Software | Developers of search tools | Prepare page images before indexing | Parsing pages before building a search index
Research | Document AI researchers | Study a data-centric parsing model | Trying the model on Chinese and English pages

People who only need plain text from digital files can skip it, since no page images are involved.

Where to use it:

• Python with the transformers library on Windows, macOS or Linux
• Pipelines that take page images and need text back
• Research setups for document parsing

Use case | Who | What they get back
Page reading | Data teams | Text read from a page image
Parsing for search | Developers | Page content in a form ready for indexing

How to set it up:

• Python on your computer
• The transformers library
• About 2.3 GB of disk for the BF16 weights

1. Install the Hugging Face hub package.
2. Download the model files.
3. Load the model with transformers, following the run steps on the model card.

pip install -U huggingface_hub

hf download opendatalab/MinerU2.5-Pro-2605-1.2B

How to use it: Send the model an image of a document page through transformers and read the text it returns. The card text provided here has no run code, so follow the full model card for the exact prompt and call.

Example: You have a scanned page in Chinese or English. You load the model, pass the page image and read the text it returns. The card text provided here shows no sample prompt or output.

For newcomers and experts: A document parsing model reads a page image the way a person would, and writes out what it finds. A good first step is to download the files and read the full card for the correct prompt, then test one clean page.

It is an image-text-to-text model in the transformers library, and the card points to a technical report for details. Because the provided text is short, treat any claim about speed or accuracy as unverified until you read that report.

Document page image
Model reads the page
Text comes back
You check the result

Good to know:

• License: Apache 2.0.
• Cost to run: none named. The steps need no API key or paid service.
• Data it reaches: it downloads files from the Hugging Face Hub, then reads the page images you give it.
• Not for: anything beyond document parsing. The card names no other uses.

On Hugging Face, 13 November 2026

• Task: image text to text
• Library: transformers
• Parameters: 1.2B
• Weights: about 2.3 GB in BF16; about 578 MB at 4-bit (estimate)
• Formats: Safetensors
• Downloads in the last 30 days: 37,179
• Likes: 115
• License: apache-2.0
• Languages: zh, en
• Quantized versions: 7
• Fine-tunes: 3
• Tags: transformers, safetensors, qwen2_vl, image-text-to-text, conversational

Links

• Model: https://huggingface.co/opendatalab/MinerU2.5-Pro-2605-1.2B
• Files: https://huggingface.co/opendatalab/MinerU2.5-Pro-2605-1.2B/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download opendatalab/MinerU2.5-Pro-2605-1.2B

Then follow the usage in its model card. Nothing in this kit runs by itself.