Nougat

Nougat (Neural Optical Understanding for Academic Documents) is an open-source model from Meta AI that converts scientific PDF documents into structured Markdown, correctly preserving mathematical equations, tables, and LaTeX formatting that standard OCR tools mangle.

Unlike general-purpose OCR, it's trained specifically on academic paper layouts, making it useful for building searchable archives of research papers or feeding paper content into downstream NLP pipelines where formula structure needs to survive the conversion.

Runs as a Python package with a pretrained model checkpoint; commonly used in research-tooling and academic-search projects rather than general document digitization.