# Baidu releases Unlimited-OCR 2 under MIT license

> The multilingual vision-language model targets document text recognition and ships with a permissive license for commercial use.

Published by The Open Weights on Oct 26, 2026. Canonical: https://theopenweights.com/news/unlimited-ocr-2-xk66

## Key facts

- Company: Baidu
- Model: Unlimited-OCR 2
- Version: 2
- Category: Vision-Language
- Modalities: Vision-Language
- License: MIT (Open weights, commercial use allowed)
- Context window: 131K tokens
- Architecture: UnlimitedOCRForCausalLM
- Precision: BF16
- Weights size: 6.7 GB
- Significance: notable
- Published: 2026-10-26
- Last verified: 2026-10-27
- Hugging Face: https://huggingface.co/baidu/Unlimited-OCR-2
- Canonical URL: https://theopenweights.com/news/unlimited-ocr-2-xk66

Baidu has published **Unlimited-OCR 2**, a vision-language model aimed at optical character recognition, on [Hugging Face](https://huggingface.co/baidu/Unlimited-OCR-2). The model is released under the MIT license, one of the most permissive options available, which clears the way for both research and commercial deployment without the usage restrictions attached to many other open releases.

As a vision-language model built for OCR, Unlimited-OCR 2 is designed to read text from images rather than generate open-ended conversation. Baidu describes the model as multilingual, suggesting it targets document understanding across more than one writing system — a practical requirement for enterprises processing invoices, forms, and scanned archives in mixed-language environments.

## Why it matters

OCR has become a quiet but critical layer in modern AI pipelines, feeding clean text into retrieval systems, translation tools, and document agents. A few points stand out:

- The **MIT license** removes friction for teams that need to fine-tune or embed the model in products.
- A **multilingual** focus broadens its usefulness beyond English-centric document workflows.
- Baidu's continued output of open models adds another option alongside the growing field of open OCR and document-parsing systems.

Baidu has not published detailed specifications such as parameter count or context length alongside this release, so teams evaluating the model will want to benchmark it against their own document sets. The weights and model card are available now on [Hugging Face](https://huggingface.co/baidu/Unlimited-OCR-2).

## Get the model

- [Hugging Face](https://huggingface.co/baidu/Unlimited-OCR-2)

## Sources

- [baidu/Unlimited-OCR-2](https://huggingface.co/baidu/Unlimited-OCR-2) — Hugging Face, Oct 26, 2026

---
Source: The Open Weights (https://theopenweights.com/). Aggregated and written by Claude, curated by humans. Cite as: "Baidu releases Unlimited-OCR 2 under MIT license", The Open Weights, Oct 26, 2026, https://theopenweights.com/news/unlimited-ocr-2-xk66