xiaomi-research / Xiaomi-OCR-0
git clone https://github.com/xiaomi-research/Xiaomi-OCR-0.git
A unified 0.8B vision-language model for document parsing, key information extraction, and OCR-centric visual question answering.
Why you might care
Useful if you need small-footprint OCR that outputs structured Markdown with HTML tables and LaTeX formulas from document pages. It is built on Qwen3.5-0.8B and includes an MCP service and a local browser demo for page and region parsing, with reported strong results on OmniDocBench and formula/table benchmarks.
+15
stars, 24h · #2,471
+690
stars, 7 days · #355
+~1.4k
stars, 30 days · #759
964
★ total · +252% this week
58
forks · 4 open issues
Star history (as recorded here)
LanguagePython
LicenseApache-2.0
Created2026-09-28 (1mo ago)
Last push10d ago
Watchers6
OwnerOrganization
Links
- GitHub: github.com/xiaomi-research/Xiaomi-OCR-0
- Homepage: https://huggingface.co/spaces/SeerRay-Lab/Xiaomi-OCR-0
- Releases · Issues · Hacker News mentions
README
Loading README…
Blurb generated by claude-haiku-5-5 7d ago. Star deltas marked ~ are extrapolated from partial history.