IT之家 · ITHome
Xiaomi open-sources 0.8B-parameter document model Xiaomi-OCR-0
Xiaomi has open-sourced Xiaomi-OCR-0, a 0.8-billion-parameter model, releasing its weights and inference code. According to Xiaomi, the model combines document parsing, key information extraction and OCR visual question answering. It handles text, tables and formulas, supports Markdown, LaTeX and JSON output, and can retrieve information, perform calculations and apply semantic reasoning across documents and charts.

What to know
- 01
Weights and inference code are available on Hugging Face and GitHub, alongside a browser demo, an MCP service and an Agent Skill toolkit.
- 02
Xiaomi says related technology is used in its “超级小爱” mobile assistant to extract information such as meal pickup codes and parcel collection codes from screens.
Why it matters
For developers, the release provides both a document-processing model and supporting tools for integration into business workflows or agent systems.
AI-assisted brief. Read the source for full details. Editorial & sourcing policy


