173 likes
Tiny 0.8B Model Beats Bigger OCR Systems on Benchmarks
What is Xiaomi-OCR-0?
Xiaomi-OCR-0 is a 0.8-billion-parameter vision-language model that can read, parse and answer questions about documents, tables and handwritten text. It runs on a single GPU and matches or exceeds larger models on standard OCR benchmarks.
Who it is for
If you need fast, accurate OCR or document understanding without massive hardware, you can drop this model into your pipeline today.