What can I use Unlimited-OCR-2 for?

Pass the scanned page images of a multi-page report in order and get one text output per page, in the same order; Give it one page image and get its text back; the page is split into tiles automatically (gundam mode); Convert a long PDF to page images with PyMuPDF, send them to an SGLang server, and get the parsed text back

Can I run Unlimited-OCR-2 on my computer?

It runs on your own hardware with a model runtime. The model card mentions Transformers, vLLM, SGLang; check its setup and hardware notes.

What do I need to use Unlimited-OCR-2?

Python 3.10 with Transformers, or a customized SGLang build for serving. Tested on an NVIDIA A800 80GB GPU; minimum hardware needs are not documented.

Unlimited-OCR-2: Turn long multi-page documents into text

Model details

Original model baidu/Unlimited-OCR-2

Understand images/documentsWrite & reason

Input Image or text Output Text

Try the browser demo Follow the model card setup

Ways to use it Browser demo · Run locally

Creator exampleSee the source example

What it does

It is an OCR model, meaning it reads text from images of pages. Its Sparse Reference Sliding Window Attention keeps a fixed-size set of working references, for example 512 tokens, so speed per step and memory stay nearly constant however long the document gets. Multi-page mode parses all pages together for cross-page context.

How to use it

Try the hosted demo on Hugging Face first. To run it yourself, load the weights with Transformers for quick tests: pass page images in order to the multi-page function, or one image to the single-page function. For serving many requests, the source recommends SGLang. It is also available on ModelScope and Baidu Cloud.

Common use cases

Pass the scanned page images of a multi-page report in order and get one text output per page, in the same order.
Give it one page image and get its text back; the page is split into tiles automatically (gundam mode).
Convert a long PDF to page images with PyMuPDF, send them to an SGLang server, and get the parsed text back.

Can I use this?

Needs a model runtime on your computer

Requirements Tested with Python 3.10, torch 2.13.0 and transformers 4.52.1. The SGLang route uses a Python 3.10 environment managed by uv, a customized SGLang package (wheel) and PyMuPDF. Downloading the weights needs no access request.

Caveat When serving with SGLang, the source warns that reserving a 0.75 memory share leaves too little headroom at 32 or more simultaneous requests.

Worth knowing Licensed under MIT. It reads page images, so PDFs are converted to images first; the SGLang setup installs PyMuPDF for that.

Why it's here

Not trending this week

Last seen on the Hugging Face trending list on 4 November 2026. It stays on BuildTube so you can still find it.

650 people have liked it on Hugging Face.
It was downloaded 390,927 times in the last 30 days.

Numbers checked on 8 November 2026; not refreshed since.

Behind it

View model page

BuildTube has not run or verified this project. Everything above is written from what the creator published.

Made this?

Quick questions

What is Unlimited-OCR-2?: It is an OCR model, meaning it reads text from images of pages.
What can I use Unlimited-OCR-2 for?: Pass the scanned page images of a multi-page report in order and get one text output per page, in the same order; Give it one page image and get its text back; the page is split into tiles automatically (gundam mode); Convert a long PDF to page images with PyMuPDF, send them to an SGLang server, and get the parsed text back
Can I run Unlimited-OCR-2 on my computer?: It runs on your own hardware with a model runtime. The model card mentions Transformers, vLLM, SGLang; check its setup and hardware notes.
What do I need to use Unlimited-OCR-2?: Python 3.10 with Transformers, or a customized SGLang build for serving. Tested on an NVIDIA A800 80GB GPU; minimum hardware needs are not documented.
What license does Unlimited-OCR-2 use?: Licensed under MIT.

More from this week

Idol Turn one picture of an object into a 3D Gaussian splat
Idol turns one photo, render or generated image of an object into a 3D Gaussian splat file.
Runs in a terminalAdded 2 Nov
P-H3-Image-Comfy Make and edit 8-megapixel images in ComfyUI
Turn short text requests into 8-megapixel images and edit pictures on your own graphics card.
Added 7 Nov
agent-desktop Let AI agents read and click desktop apps
For developers building computer-use agents: agent-desktop lets an AI agent read and operate desktop apps on Mac, Windows or Linux from a command line.
+13,858 stars this weekAdded 3 Nov

Related projects

Unlimited-OCR Download and run the Baidu OCR model on page images and PDFs
Same topic: transformers
Needs a model setupAdded 21 Oct
TeleOCR Turn document scans and photos into structured text
Same topic: transformers
Needs a model setupAdded 28 Sept
NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 Ask about documents and video
Same topic: transformers
Needs a model setupAdded 18 Oct
NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16 Ask about images and video
Same topic: transformers
Needs a model setupAdded 20 Oct
MiniCPM-OCR-1.0 Turn scans and PDFs into editable text (1.3B)
Same topic: safetensors
Needs a model setupAdded 31 Oct
Qwen3.8-27B Run a 27B model that reads images and video yourself
Same topic: transformers
Needs a model setupAdded 25 Sept

Explore by topic

Post this

For makers

Made this? Show it’s on BuildTube — add the badge to your README or embed the card on your site.

README badge (Markdown)

[![Featured on BuildTube](https://buildtube.app/badge/hf-baidu-unlimited-ocr-2.svg)](https://buildtube.app/p/hf-baidu-unlimited-ocr-2/)

Embed card (HTML)

<iframe src="https://buildtube.app/embed/hf-baidu-unlimited-ocr-2/" width="560" height="240" style="border:0;max-width:100%" loading="lazy" title="Unlimited-OCR-2 on BuildTube"></iframe>