SoMark open-sources OXR-1.0, a 1.3B document parser

Chinese startup SoMark has released OXR-1.0, a 1.3-billion-parameter model that converts PDFs and document images into Markdown or JSON with layout coordinates. It ranks first on three public benchmarks: 97.18 on OmniDocBench, 87.84 on olmOCR-Bench and 87.76 on DocParsingBench.

Pandaily

Open source

Code and weights are on GitHub, Hugging Face and ModelScope, with a demo and free API
On an RTX PRO 6000 it parsed 100 OmniDocBench pages in about 101 seconds vs 145 for PaddleOCR-VL-1.6
The model recognizes 13 element types, including tables, formulas and chemical structures
Its layout module is independent and swappable, and post-training needs no manual labeling

#SoMark#OXR

SoMark open-sources OXR-1.0, a 1.3B document parser

Jina AI Releases jina-ocr-v1: A 3.4B MoE Document Parser With Built-In Speculative Decoding for Low-Budget GPUs

Baidu open-sources Unlimited OCR 2 for hundred-page document parsing

MemT-1.0 brings native memory into the model, delivering 1.86x faster inference

MOSS-Image Passes 1.3 Million Downloads in Four Days, Tops Hugging Face Trends