Nemotron 3.5 Parse

Nemotron 3.5 Parse turns document-page images into structured, agent-ready content. It extracts text in reading order and captures layout details such as tables, charts, and formulas, helping make complex enterprise documents easier to use in retrieval and AI workflows.

Use Parse to prepare scanned or image-based documents for search, RAG, and downstream analysis. It takes a page image and task prompt as input, and returns text with layout information, including Markdown and bounding boxes.

See Hugging Face model card