Baidu launches open-source unlimited ocr 2 model for document parsing

On October 30, 2026, IT Home reported that Baidu officially released and open-sourced the Unlimited OCR 2 model on October 27. The model significantly enhances document parsing capabilities by removing page limits, allowing for the processing of over 3,600 pages of the entire Harry Potter series, which consists of seven books. The model supports native page element segmentation and recognition, along with cross-page text merging and table recognition merging, enabling faster and more efficient processing with the same computational power. Baidu stated that the Unlimited OCR 2 achieved scores of 84.6 on the MPDocBench benchmark for multi-page document parsing and 77.1 on the MarathonDocBench benchmark for ultra-long document parsing, achieving state-of-the-art performance in end-to-end processing.