# Tencent Releases 1B Parameter HunyuanOCR Model

> The new vision-language model from Tencent Hunyuan offers a compact, end-to-end solution for optical character recognition.

Published by The Open Weights on Nov 18, 2025. Canonical: https://theopenweights.com/news/hunyuanocr-bksf

## Key facts

- Company: Tencent
- Model: HunyuanOCR
- Version: 1.0
- Category: Vision-Language
- Modalities: Vision-Language
- License: Other (Open weights, commercial use allowed)
- Parameters: 1B
- Context window: 131K tokens
- Architecture: HunYuanVLForConditionalGeneration
- Precision: BF16
- Weights size: 2.2 GB
- Significance: notable
- Published: 2025-11-18
- Last verified: 2026-06-17
- Hugging Face: https://huggingface.co/tencent/HunyuanOCR
- Canonical URL: https://theopenweights.com/news/hunyuanocr-bksf

Tencent has released HunyuanOCR, a new vision-language model specialized for reading text in images. At a relatively compact one billion parameters, the model provides an efficient, open-source tool for developers working on optical character recognition (OCR) tasks.

HunyuanOCR uses an end-to-end architecture, which simplifies the traditional OCR pipeline. Instead of first detecting text boxes and then separately recognizing the characters inside them, the model processes the entire task in a single step. This integrated approach can improve performance on challenging inputs like dense documents or text in natural scenes.

The model's capabilities are suited for a range of applications, including document digitization, extracting information from forms, and reading text from real-world photos like street signs or product labels. All model assets are available on the [Hugging Face Hub](https://huggingface.co/tencent/HunyuanOCR) under a permissive Apache 2.0 license, encouraging both research and commercial use.

This release from the Tencent Hunyuan team reflects a growing industry trend of releasing smaller, specialized models. While massive general-purpose models attract headlines, focused tools like HunyuanOCR provide a practical and efficient solution for developers needing to solve a specific, common problem.

## Get the model

- [Hugging Face](https://huggingface.co/tencent/HunyuanOCR)

## Sources

- [tencent/HunyuanOCR](https://huggingface.co/tencent/HunyuanOCR) — Hugging Face, Nov 18, 2025

---
Source: The Open Weights (https://theopenweights.com/). Aggregated and written by Claude, curated by humans. Cite as: "Tencent Releases 1B Parameter HunyuanOCR Model", The Open Weights, Nov 18, 2025, https://theopenweights.com/news/hunyuanocr-bksf