1. Tutorial Introduction

DeepSeek-OCR, released by DeepSeek Inc. in October 2025, represents a preliminary study on the feasibility of compressing long contexts from images. DeepEncoder, the core engine, aims to maintain low activation levels while achieving a high compression ratio under high-resolution input, ensuring the number of visual tokens remains within a controllable and optimized range. Experiments show that when the number of text tokens does not exceed 10 times the number of visual tokens (i.e., compression ratio < 10×), the model achieves a decoding (OCR) accuracy of 971 TP3T. Even at a compression ratio of 20×, the OCR accuracy remains approximately 601 TP3T. This demonstrates considerable promise for research directions such as long context compression of historical documents and memory decay mechanisms in large models. The related paper is titled "...".DeepSeek-OCR: Contexts Optical Compression".

This tutorial uses a single RTX 5090 graphics card as the default resource, but a minimum single RTX 4090 graphics card can be used to start the program.

2. Project Examples

3. Operation steps

1. After starting the container, click the API address to enter the Web interface

2. After entering the webpage, you can upload images and parse text.

If "Bad Gateway" is displayed, it means the model is initializing. Since the model is large, please wait about 2-3 minutes and refresh the page.

How to use

 3. Output Results 

4. Discussion

🖌️ If you see a high-quality project, please leave a message in the background to recommend it! In addition, we have also established a tutorial exchange group. Welcome friends to scan the QR code and remark [SD Tutorial] to join the group to discuss various technical issues and share application effects↓

Citation Information

The citation information for this project is as follows:

@article{wei2025deepseek,
  title={DeepSeek-OCR: Contexts Optical Compression},
  author={Wei, Haoran and Sun, Yaofeng and Li, Yukun},
  journal={arXiv preprint arXiv:2510.18234},
  year={2025}
}