Baidu open-sourced Unlimited OCR, a 3B-parameter end-to-end model for long document parsing. It activates ~570M parameters and features Reference Sliding Window Attention for continuous multi-page parsing with constant KV Cache. It scored 93.92% on OmniDocBench v1.6, and is 12.7-35% faster than DeepSeek OCR, reaching 10k GitHub stars in 5 days.





Home
