Xiaomi open-sources Xiaomi-OCR-0, a 0.8B document understanding model
Overview
Xiaomi has open-sourced Xiaomi-OCR-0, a 0.8-billion-parameter model for document parsing, key information extraction and OCR visual question answering, with weights and inference code on Hugging Face and GitHub.
The model runs three task types in a single model and comes with a browser demo, an MCP service and an Agent Skill toolkit for developers.
Xiaomi reports a score of 96.83 on OmniDocBench v1.6, which it says ranks first among models under 1 billion parameters and second overall. The company's own benchmark figures are not independently verified. Xiaomi also says the model is built on Qwen3.5-0.8B-Base, was trained on about 170 million samples, and already runs in its Super XiaoAi assistant.
Written by AI from the articles below · updated Oct 30, 11:29 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- PandailyNewsXiaomi open-sources Xiaomi-OCR-0, a 0.8B document model for parsing and extraction
AIXiaomi has open-sourced Xiaomi-OCR-0, a 0.8 billion parameter model for document parsing, key information extraction and OCR visual question answering, with weights and inference code on Hugging Face and GitHub. The company reports 96.83 on OmniDocBench v1.6, first among models under 1 billion parameters and second overall, though all figures are Xiaomi's own. The model is built on Qwen3.5-0.8B-Base and trained on about 170 million samples, and Xiaomi says it already runs in its Super XiaoAi assistant.
Heat Trend
Not enough continuous observations to show a trend yet.