Xiaomi open-sources a 0.8B document-reading model
Xiaomi-OCR-0 handles parsing, extraction and OCR questions in one small model.

Xiaomi has released Xiaomi-OCR-0, an open-source document understanding model with 0.8 billion parameters.
It combines document parsing, key information extraction and OCR-focused visual question answering. Xiaomi reports a score of 95.24 on Real5-OmniDocBench and says it ranks first in document parsing on OmniDocBench v1.6 among models under 1B parameters.
Model weights, inference code, an interactive demo, an MCP service and an agent skill are open source.




