Developers building OCR pipelines on their own hardware get a free Apache-licensed model, though you'll need a GPU to run it.
Kenning's take
What it does
DeepSeek-OCR 2 is an open vision-language model for reading text from documents and images, including converting pages to Markdown. Its new DeepEncoder V2 reorders visual tokens by the layout of the page rather than a fixed scan order. Code and weights are free to download under the Apache 2.0 licence.
- Document to Markdown conversion
- DeepEncoder V2 that reorders visual tokens by page layout
- Runs with vLLM or Hugging Face Transformers
- About 3.4 billion parameters
- Apache 2.0 licence
Who it suits. Developers building document parsing, OCR or data extraction pipelines who can run their own models.
Key facts
- Category
- AI models
- Part of
- DeepSeek
- Made by
- DeepSeek
- Released
- 27 January 2026
- How to get it
- free download from GitHub and Hugging Face; needs a GPU to run
- Works on
- open weights
- In the UK
- Yes
- Free plan
- open weights, Apache 2.0
FAQ
When was DeepSeek-OCR 2 released?
The Hugging Face model page was created on 27 January 2026 and the paper was submitted on 28 January 2026.
Is DeepSeek-OCR 2 free?
Yes. The code and weights are published under the Apache 2.0 licence.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
DeepSeek V4.1 Flash
DeepSeek's efficient open-weight model that reads images and text, with a one million token context
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.