MongoDB - Build AI That Scales
Mini toolsdots.ocr
dots.ocr is a single vision-language model that jointly learns layout detection, text recognition, tables, formulas and reading order instead of using multi stage OCR pipelines. Built on a compact 1.7B LLM, it reaches SOTA on OmniDocBench and strong performance on multilingual internal benchmarks, supporting over 100 languages. The project also introduces XDocParse, a 126 language benchmark where dots.ocr sets a strong baseline, showing that a unified VLM can rival or beat specialized detectors.
New
Text
Gen 7
Overview
dots.ocr is a 1.7B parameter vision-language model for multilingual document layout parsing, unifying layout detection, OCR and reading order in one model, and achieving state-of-the-art results on OmniDocBench.
About Rednote HiLab
Founded in Shanghai in 2013, rednote is a platform where users capture and share their lives through photos, text, videos, and live streams, building an interactive community around shared interests. Guided by its mission to Inspire Lives, rednote is becoming a vibrant hub for diverse lifestyles and a trusted companion to millions.
Tools using dots.ocr
No tools found for this model yet.
Last updated: February 25, 2026
