PaddleOCR license
and what it allows.
PaddleOCR code and the PaddleOCR-VL weights are released under Apache-2.0, which allows commercial use. GLM-OCR weights are MIT. Here is what each license asks of you, and what changes when you use a hosted API.
Licenses for PaddleOCR, GLM-OCR, and their layout model
Published licenses for each component of the two open-weight OCR pipelines.
| Component | License | What it is |
|---|---|---|
| PaddleOCR (code) | Apache-2.0 | The PaddlePaddle/PaddleOCR repository and pipelines. |
| PaddleOCR-VL (weights) | Apache-2.0 | The vision-language model that reads each layout region. |
| PP-DocLayoutV3 (weights) | Apache-2.0 | Layout detection used by both PaddleOCR-VL and GLM-OCR pipelines. |
| GLM-OCR (weights) | MIT | Z.ai document OCR model, hosted on OpenParser as glm-ocr. |
| glmocr SDK | Apache-2.0 | Z.ai pipeline code that runs layout detection and region recognition. |
What the licenses ask of you
Both allow commercial use. The conditions apply when you redistribute the code or weights.
Apache-2.0 asks you to
- Keep the license text and any NOTICE file when you redistribute the code or weights.
- Mark files you changed.
- Not use the project names or trademarks to endorse your product.
MIT asks you to
- Keep the copyright notice and license text in copies you redistribute.
- Accept that the software comes without warranty.
PaddleOCR license FAQ
Commercial use, GLM-OCR, hosted APIs, and dependencies to check when self-hosting.
Can I use PaddleOCR commercially?
Yes. PaddleOCR and the PaddleOCR-VL weights are released under Apache-2.0, which permits commercial use, modification, and distribution, and includes a patent license from contributors. Keep the license and notices when you redistribute.
What license is GLM-OCR under?
The GLM-OCR weights are MIT licensed. Its recommended pipeline also uses PP-DocLayoutV3 for layout, which is Apache-2.0.
Do I need to show a license if I call a hosted OCR API?
Calling an API does not redistribute the model to you, so the redistribution terms apply to whoever runs it. On OpenParser, we run PaddleOCR-VL and GLM-OCR and keep their licenses with our deployment. Your obligations come from the OpenParser terms of service.
Are there copyleft dependencies to watch for when self-hosting?
Check every dependency in the pipeline as well as the model. For example, the GLM-OCR SDK imports PyMuPDF, which is AGPL-3.0. Our hosted deployment does not install PyMuPDF and rasterizes PDFs with pypdfium2 instead.
Is this legal advice?
No. It summarizes the published licenses as of October 2026. Read the license files in each repository and ask your counsel for decisions about your product.
Use PaddleOCR-VL or GLM-OCR through one API
OpenParser runs both open-weight models at $1 per 1,000 pages. New accounts get $10 in free credits.