Reportedly: Xiaomi Open-Sources Xiaomi-OCR-0, a 0.8B Document Model That Tops OmniDocBench in Its Size Class
Model: Xiaomi-OCR-0 · availability, license and releases
Reported by Pandaily · not yet confirmed by the company or a second independent outlet. We update this page when it is.
Xiaomi open-sourced Xiaomi-OCR-0, a 0.8B model for document parsing, key-field extraction and OCR question answering. It scores 96.83 on OmniDocBench v1.6, first among models under 1B. The source release was published on 2026-10-31, Pandaily reported.
China context
- Original name
- 小米
- Outside China
- Open weights · pandaily.com
- Claims
- Company-reported; not yet independently evaluated
- For builders
- A 0.8B open-weights OCR model can be fine-tuned and deployed on modest hardware, enabling custom document extraction without large GPU clusters.
- For investors
- Xiaomi's open-weights release signals a push into developer-facing AI tools, potentially expanding its ecosystem beyond consumer hardware.
Xiaomi has released Xiaomi-OCR-0 as open weights. The 0.8B model handles document parsing, key-field extraction, and OCR question answering. On OmniDocBench v1.6 it achieves 96.83, ranking first among models under 1B parameters.
The model's 96.83 OmniDocBench v1.6 score is company-reported. No independent evaluation is provided in the evidence. The benchmark ranking is limited to models under 1B parameters.
Developers outside China can now use a small, open-weights document OCR model that claims top performance in its size class, reducing the cost of building document-processing pipelines without relying on closed APIs.
Open weights allow enterprises to self-host document parsing and key-field extraction, potentially lowering per-document costs and avoiding data-sharing with third-party APIs.
Next verifiable signals: independent benchmark reproductions, community fine-tunes, and whether Xiaomi releases larger OCR models or commercial API access.