OpenDoc-0.1B: A Lightweight Revolution in Document Parsing with 0.1B Parameters

OPENDOC-0.1BDOCUMENT PARSINGLIGHTWEIGHT AIOMNIDOCBENCHTEXT RECOGNITIONFORMULA RECOGNITIONTABLE RECOGNITIONEFFICIENCYCOST REDUCTIONPADDLEOCR
OpenDoc-0.1B, developed by Fudan University's Visual and Learning Lab, challenges the trend of massive AI models by achieving high performance (90.57% on OmniDocBench v1.5) with only 100 million parameters, demonstrating the potential of efficient, specialized models for document parsing tasks.
Generated from mp.weixin.qq.com
Mar 3, 2026

Scroll to Zoom • Drag to Pan