Xiaomi open-sources 0.8B document model Xiaomi-OCR-0
Xiaomi has open-sourced a lightweight document model called Xiaomi-OCR-0. The weights and code are now live on Hugging Face and GitHub. With just 0.8B parameters, one model handles document parsing, key info extraction, and OCR visual Q&A.
It is built on Qwen3.5-0.8B-Base and trained with about 170 million samples. It reads text, tables, and math formulas, and pulls fields from invoices and receipts. It scored 96.83 on the OmniDocBench v1.6 benchmark, ranking first among models under 1B.
The tech already powers Super XiaoAi on Xiaomi phones, grabbing delivery codes and bill info from the screen. Xiaomi also offers a browser demo, an MCP service, and an Agent Skill toolkit for developers.
A universal smart platform that brings multimodal AI, agentic intelligence, and natural voice synthesis together, so machines can see, hear, understand, and act alongside you.
ByteDance AI chatbot development platform with LLM integration, plugin ecosystem and cross-platform deployment to Doubao and Feishu
Comments
Loading comments…