Skip to content
Live
Trending storyPast story

Baidu Open Sources Unlimited OCR 2 Long Document Model

2 reports2 outlets20 days agoUpdate

The story so far

AI overview

Baidu has released and open-sourced Wenxin Unlimited OCR 2, an OCR visual language model with 3.34B parameters, with weights released under the MIT license and code hosted on GitHub. Baidu claims it can parse documents totaling over 3,600 pages in a single request, natively supporting layout element segmentation recognition, cross-page text merging, and table recognition merging. In terms of operation, the model requires one CUDA GPU under BF16 precision, with Baidu testing using an NVIDIA A800-80GB. Loading requires setting trust_remote_code=True, which will execute the Python code in the repository; SGLang deployment uses a custom wheel provided by Baidu, not the upstream version. Baidu's self-reported test results show that the model scored 84.6 on MPDocBench and 77.1 on MarathonDocBench, and was ranked first among 9 other models on both of its two multi-page benchmarks.

AI-written from the reports · Updated 20 days ago

Coverage timeline

Follow the coverage to see every side of the story.

30 Oct
  1. ITHome · AI
    Baidu open-sources Wenxin Unlimited OCR 2 document parsing model, supporting 'infinite' number of pages

    Baidu has open-sourced the Wenxin Unlimited OCR 2 model, which supports parsing documents totaling over 3,600 pages at once, natively supporting layout element segmentation recognition and cross-page text merging, table recognition and merging. It scored 84.6 on the MPDocBench benchmark test and 77.1 on the MarathonDocBench benchmark test.

27 Oct
  1. MarkTechPostFeatured
    Baidu Open-Sources Unlimited OCR 2 for Hundred-Page Document Parsing

    Baidu open-sources Unlimited OCR 2, a 3.34B parameter OCR visual language model that can parse hundreds of pages of books, financial reports, and reports in a single request. The weights are released under the MIT license, and the code is on GitHub.