0.8B! Xiaomi open-sources new model, enabling Super Xiao Ai to recognize screens and proactively handle tasks

What happened: Xiaomi released and open-sourced Xiaomi-OCR-0, a 0.8B parameter document understanding model that combines document parsing, key information extraction, and visual question answering. Xiaomi is publishing the model weights and inference code. Key numbers: 0.8B parameters. Why it matters: The content states the model nears Qwen3.5-4B on VQA at about 20% of its parameter scale, and that open weights plus Agent tooling may lower the barrier to integrating document understanding on devices and in agents.

Xiaomi released and open-sourced Xiaomi-OCR-0, a 0.8B parameter document understanding model that combines document parsing, key information extraction, and visual question answering. Xiaomi is publishing the model weights and inference code.

The content states the model nears Qwen3.5-4B on VQA at about 20% of its parameter scale, and that open weights plus Agent tooling may lower the barrier to integrating document understanding on devices and in agents.