High-efficiency models for optical character recognition and document understanding.

Model Details

Variants

Name Parameters Notes
DeepSeek-OCR Released Oct 2025
DeepSeek-OCR-2 Released Feb 2026
visionopen-weight