Compare software

Marker vs PaddleOCR: catalog facts
MarkerPaddleOCR
Free

The code is free under Apache 2.0; model weights have a separate licence, and Datalab…

Free

Free and open source under the Apache licence.

Self-hosted, Command line Windows, macOS, Linux, Command line
  • Handles tables, equations and code blocks
  • Accepts many input formats besides PDF
  • Parses layout, tables and formulas, not just lines of text
  • Outputs JSON and Markdown ready for downstream processing
  • Model weights are under a separate licence from the code
  • Needs a Python environment and benefits from a GPU
  • Built on the PaddlePaddle framework, which is a large dependency
  • Much of the documentation and community is in Chinese
Downloadable app, Self-hosted Downloadable app
Open source Open source
License: Apache-2.0 License: Apache-2.0
No account needed Not stated if an account is needed
Not stated if it works offline Not stated if it works offline

Checked September 23, 2026

Checked September 22, 2026

Catalog facts only. Anything “not stated” is unconfirmed. Check full listings for details.