TOOLORA / Local reading edition
MinerU
Parse complex documents with OCR and vision-language models, retaining page/block locators.
Overview
MinerU 4 offers a CLI, Python SDK and local services for document parsing and agent reading. PDFs and images support four tiers combining native parsing, OCR and vision-language models. Office formats use local Flash parsing, not necessarily a VLM. Page/block locators support continued reading and citations. It is a document processor, not a ready-made RAG platform; remote parsing requires explicit configuration.
Usage guide & costs
- Before you start
The Apache-2.0-based license adds conditions: combined affiliate MAU over 100 million or monthly revenue over US$20 million requires a commercial license. Third-party online services must prominently credit MinerU. Check model/dependency licenses separately.
Official details ↗ · Official details ↗
Official sources and references
- Official websitehttps://github.com/opendatalab/MinerU
- Documentationhttps://opendatalab.github.io/MinerU/
- Documentationhttps://opendatalab.github.io/MinerU/usage/tiers/
- Documentationhttps://github.com/opendatalab/MinerU/blob/master/README.md
Descriptions summarize public information; consult official websites for features, pricing and regional availability.