deepseek-ocr:latest

Upstream reference: DeepSeek-OCR

Publisher: DeepSeek

DeepSeek-OCR is a vision-language research model for optical text compression and converting document images into text or structured Markdown.

Publisher-described context, separate from Sigma5C observations and recorded catalog statistics.

Published upstream specifications

Input/output role

Document images to text Source

Upstream license label

MIT Source

Publisher-described uses

• OCR and document-to-Markdown extraction. Source

Reference limitations

• Image-size and crop settings are application choices; document extraction can contain errors and does not establish factual correctness. Source

A recorded tag and an upstream card association do not verify the deployed checkpoint, quantization, configured context or current hosted availability.

Publisher sources

• DeepSeek-OCR official documentation — DeepSeek. Checked 2026-11-02.

Links

Recorded references do not establish downloadable weights, licensing or current availability.

• DeepSeek — official site
• Ollama
• Hugging Face
• GitHub (maker)

Where deepseek-ocr:latest comes from

Named in 2 observation pages

• 14B-32B local models — all 47 · V1
• Models as data workers · V1