# Unlimited OCR 2



Baidu



Baidu open-sourced Unlimited OCR 2 for document parsing. Its SR-SWA mechanism condenses text-visual attention into a fixed set of working references, so it parses documents of almost any length with near-constant decoding latency and memory. Baidu reports SOTA on MPDocBench and MarathonDocBench.


## Availability

Open weights under the MIT license, about 3.3B parameters; Hugging Face baidu/Unlimited-OCR-2, code at GitHub baidu/Unlimited-OCR.
- Released: Oct 27, 2026 10:41 · Beijing Time (UTC+8)
- Review updated: Oct 27, 2026 18:07 Tue · Beijing Time (UTC+8)

## Sources

- Source excerpt: https://x.com/Baidu_Inc/status/2114910381311512987
- Source excerpt: https://huggingface.co/baidu/Unlimited-OCR-2

Canonical entry: https://www.ailyre.com/source/?id=baidu-unlimited-ocr-2&lang=en