MinerU
MinerU is a scientific document parser: drop a paper in the Workbench (or pip the repo) and get Markdown, HTML, JSON, and LaTeX for formulas.
OUR TAKE
Pick MinerU for research and RAG teams who will run the repo or stay on hosted Free/Pro. Free has no API.
Good fit
Research and RAG teams who will run the repo or stay on hosted Free/Pro.
Poor fit
AP teams who need invoice key-values (Parseur).
Self-host · Pro · Monthly $9.90/mo · Pro · Annual $59.90/mo
USD, before tax · Checked 2026-10-28 · mineru.io
OUR TAKE
Pick MinerU for research and RAG teams who will run the repo or stay on hosted Free/Pro. Free has no API. Five pages/day is a poke, not a pipeline.
Good fit
Research and RAG teams who will run the repo or stay on hosted Free/Pro.
Poor fit
AP teams who need invoice key-values (Parseur).
Self-host · Pro · Monthly $9.90/mo · Pro · Annual $59.90/mo
USD, before tax · Checked 2026-10-28 · mineru.io
What it costs
USD, before tax · Checked 2026-10-28 · mineru.io
Self-host
under the project license. Separate commercial terms above 100M MAU or M monthly revenue; online services must attribute MinerU
$0
Hosted Free
5 pages every day. Files 10 MB; images 5 MB; PDFs 5 pages. Web only — no API
$0
Pro · Monthly
1,000 pages/month. Files 200 MB; PDFs 500 pages. API included
Lowest paid
$9.90/mo
Pro · Annual
Same 1,000 pages/month + API
$59.90/mo
Pro also keeps the daily free allowance. Monthly pages do not roll. Outputs kept 7 days.
Not published by the vendor
Not published by the vendor
Checked 2026-10-28·mineru.io
Before you pay
Good fit
Research and RAG teams who will run the repo or stay on hosted Free/Pro.
Poor fit
AP teams who need invoice key-values (Parseur).
Watch out
Free has no API. Five pages/day is a poke, not a pipeline.
Instead, try
Docling if MIT local parse when you want IBM’s DoclingDocument; PaddleOCR if Apache VL OCR on Hugging Face
Compare MinerU with
Limits, in full
- 01
Free has no API. Five pages/day is a poke, not a pipeline.
- 02
mineru.net KIE is sunsetting 2026-06-10. Use mineru.io / the repo.
- 03
A science parser, not Textract Forms
If MinerU isn't it
Why: MIT local parse when you want IBM’s DoclingDocument.
Free· open source,MIT
You bring hardwareChecked 2026-10-28·github.com/docling-project/docling
vs MinerU
Why: Apache VL OCR on Hugging Face.
Free· open source,Apache-2.0
The Space can be slowChecked 2026-10-28·github.com/PaddlePaddle/PaddleOCR
vs MinerU
Why: connector ETL at $0.015 / page.
Free· paid from$0.015
$0.015/page after 10kChecked 2026-10-22·unstructured.io/pricing
vs MinerULimits & alternatives3 limits · Docling, PaddleOCR, Unstructured instead
Limits, in full
- 01
Free has no API. Five pages/day is a poke, not a pipeline.
- 02
mineru.net KIE is sunsetting 2026-06-10. Use mineru.io / the repo.
- 03
A science parser, not Textract Forms
If MinerU isn't it
Why: MIT local parse when you want IBM’s DoclingDocument.
Free· open source,MIT
You bring hardwareChecked 2026-10-28·github.com/docling-project/docling
vs MinerUWhy: Apache VL OCR on Hugging Face.
Free· open source,Apache-2.0
The Space can be slowChecked 2026-10-28·github.com/PaddlePaddle/PaddleOCR
vs MinerUWhy: connector ETL at $0.015 / page.
Free· paid from$0.015
$0.015/page after 10kChecked 2026-10-22·unstructured.io/pricing
vs MinerUPricing notes & what changed3 notes · 2 sources
What it is for
MinerU is a scientific document parser: drop a paper in the Workbench (or pip the repo) and get Markdown, HTML, JSON, and LaTeX for formulas. Use it when the job is papers and chem drawings → RAG-ready files, not an invoice mailbox.
The software is open source (project license $0 to run). Hosted mineru.io adds a Free and Pro card.
What makes it unique: the cover is MinerU Online’s Workbench — General documents / Chemical structures tabs, Choose file / Enter link, and a parsed chemistry figure in the laptop pane — not a logo tile.
Key features
- Workbench — upload or paste a URL; Force OCR toggle (mineru.io).
- Markdown / JSON / LaTeX — tables, formulas, and reading order.
- Chemical structures — molecule and reaction extraction on the same .
- Self-host — OpenDataLab/MinerU repo (commercial license needed above huge MAU/revenue; online services must attribute).
- Hosted API — Pro only (developer).
Who it's for
Best fit: research and RAG teams who will run the repo or stay on hosted Free/Pro.
Poor fit: AP teams who need invoice key-values (Parseur).
Pricing
Verified 2026-10-28 from mineru.io pricing cards and github.com/opendatalab/MinerU.
- Self-host — $0 under the project license. Separate commercial terms above 100M MAU or $20M monthly revenue; online services must attribute MinerU.
- Hosted Free — $0 / month. 5 pages every day. Files 10 MB; images 5 MB; PDFs 5 pages. Web only — no API.
- Pro · Monthly — $9.90 / month. 1,000 pages/month. Files 200 MB; PDFs 500 pages. API included.
- Pro · Annual — $59.90 / year. Same 1,000 pages/month + API.
- Pro also keeps the daily free allowance. Monthly pages do not roll. Outputs kept 7 days.
Limitations / costs
- Free has no API. Five pages/day is a poke, not a pipeline.
- mineru.net KIE is sunsetting 2026-06-10. Use mineru.io / the repo.
- A science parser, not Textract Forms.
Alternatives
- Docling — MIT local parse when you want IBM’s DoclingDocument.
- PaddleOCR — Apache VL OCR on Hugging Face.
- Unstructured — connector ETL at $0.015 / page.
Official tutorial
No official first-party how-to YouTube found as of 2026-10-28. Written path: mineru.io plus the MinerU README.
Website
Source
Catalog row curated for AI Tools Nav. Cover is MinerU Online’s Workbench. Dollars from mineru.io (Free $0 / 5 pages/day; Pro $9.90 / mo or $59.90 / yr).
Last verified
2026-10-28 editorial check.