HunyuanOCR

Tencent's 1.1B end-to-end OCR model for parsing, spotting, extraction and image translation. Not licensed in the EU, UK or South Korea.

In short. HunyuanOCR is a small AI model from Tencent that reads text from photos, scans and PDFs and can also pull out fields, locate text and translate text inside images. It runs on a single GPU or a laptop. Its license, however, does not cover the EU, the UK or South Korea.

Summary

Tencent released HunyuanOCR 1.0 on 25 November 2025 and HunyuanOCR-1.5 on 7 July 2026; the Hugging Face repository now holds 1.5 at its root, with the old 1.0 archived in a subfolder. It is a single end-to-end vision-language model of about 1.1 billion parameters, which is about 2.2 GB in BF16. It handles document parsing to Markdown, HTML tables and LaTeX formulas, text spotting with positions, information extraction, text-image translation and multi-image questions. Version 1.5 does not change the backbone. It raises the maximum image resolution to 4K and the context to 128K tokens, and adds training data for low-resource languages, ancient scripts and chart and table parsing.

The main addition is DFlash speculative decoding: a light block-diffusion draft model proposes several tokens at once and the main model verifies them, which keeps the output unchanged. Tencent's report gives a 6.37 times speedup with Transformers and 2.14 times with vLLM, and says the model ranks among the top end-to-end OCR systems on OmniDocBench v1.6. These are Tencent's own numbers. A GGUF route through llama.cpp supports CPUs, consumer GPUs and laptops. The unified inference environment needs CUDA 13 and vLLM 0.25.1 or newer.

The license is the Tencent Hunyuan Community License, not Apache or MIT. It does not apply in the European Union, the United Kingdom or South Korea, requires a separate license above 100 million monthly active users, and bars using the model or its outputs to improve other AI models. Read it before building a product on HunyuanOCR, and compare it with PaddleOCR-VL or GLM-OCR if you need a permissive license.

Good for

  • Document parsing, text spotting and field extraction in one small model
  • Translating text inside images, such as menus, posters and screenshots
  • Low-resource and ancient-script OCR, where Tencent added targeted data
  • Local deployment on a laptop or consumer GPU through llama.cpp

Watch out for

  • The license excludes the EU, the UK and South Korea, and forbids using outputs to improve other AI models
  • The speed and ranking claims come from Tencent's own report
  • The unified environment needs CUDA 13, and the DFlash draft model ships in a subfolder of the repository
  • The old 1.0 checkpoint sits under the v1.0 subfolder, so download the right files

Performance claims come from the developer and have not been independently verified.

How much memory does it need?

Estimates for the HunyuanOCR-1.5 variant. A mixture-of-experts model must hold all of its experts in memory even though only some are used for each token, so total parameters decide the memory and active parameters decide the speed.

Variants Parameters BF16 FP8 / INT8 4-bit (Q4) Fits on
HunyuanOCR-1.5 BF16 1.12B 2.58 GB 1.29 GB 0.773 GB Laptop or phone

Estimates: weights plus about 15% overhead. Real usage grows with context length and depends on the software.

VRAM calculator

The license in plain words

A royalty-free license with real limits. It is valid only outside the European Union, the United Kingdom and South Korea, large services with over 100 million monthly users must ask Tencent first, and you may not use the model or its outputs to improve other AI models.

  • Attribution or model name must be shown
  • Free commercial use only below a revenue cap
  • Not licensed in some regions
  • Acceptable-use policy applies

More about Tencent Hunyuan

See also

Frequently asked questions

What is HunyuanOCR?
HunyuanOCR is a small AI model from Tencent that reads text from photos, scans and PDFs and can also pull out fields, locate text and translate text inside images. It runs on a single GPU or a laptop. Its license, however, does not cover the EU, the UK or South Korea.
Can I use HunyuanOCR commercially?
Yes, with conditions. HunyuanOCR is released under the Tencent Hunyuan, which allows commercial use but adds extra rules: Attribution or model name must be shown; Free commercial use only below a revenue cap; Not licensed in some regions; Acceptable-use policy applies. Read the license text before you ship a product.
Is HunyuanOCR open source?
HunyuanOCR is open-weight, not open source in the strict sense. You can download the model files, but the Tencent Hunyuan adds conditions that the Open Source Initiative definition does not allow.
Can I run HunyuanOCR on my own computer?
Yes. The smallest variant, HunyuanOCR-1.5, needs about 0.773 GB of memory at 4-bit precision, which fits a modern laptop or a mid-range graphics card.
When was HunyuanOCR released?
HunyuanOCR was released by Tencent on July 7, 2026.
How long a context window does HunyuanOCR have?
HunyuanOCR supports a context window of up to 128K tokens.

Official links

Sources

Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.

Release date and facts last verified November 11, 2026.

Related models

Model Released Parameters Context License Smallest variant needs
PaddleOCR-VL Baidu 960M – Apache 2.0 (Permissive) 0.662 GB Laptop or phone
GLM-OCR Z.ai (Zhipu AI) 900M – MIT (Permissive) 0.621 GB Laptop or phone
MiniCPM-OCR 1.0 OpenBMB 1.3B 256K Apache 2.0 (Permissive) 0.897 GB Laptop or phone
Xiaomi-OCR-0 Xiaomi 873M 256K Apache 2.0 (Permissive) 0.602 GB Laptop or phone