MinerU
Document layout, reading order, figures, and structured content.
Planned version 3.4.5 · Up to 500 pages
THE DOCUMENT WORKBENCH
Your PDF.
Three ways to
make it useful.
Turn a document into text, tables, and structured data. Choose your processors, compare their output, and keep everything in one place.
Account required · PDFs only · For US residents age 18+
ONE DOCUMENT. YOUR CHOICE.In development
Start with your document
Sign in before choosing a file
01
MinerU
MD + JSON500 pages · 50 MiB maximum
02
Docling
MD + JSON500 pages · 50 MiB maximum
03
Mistral OCR
MD + JSON100 pages · 50 MiB maximum
Choose one, two, or all three
Separate results. One account.ZIP
01 / CHOOSE
One upload, your engines.
Select the processors that fit your document. Each returns its own labeled set of files.
02 / REVIEW
See the price before paying.
Review each engine’s price and the subtotal, then pay once. Launch prices are still being verified.
03 / RETURN
Your results, for 30 days.
Return to your account to compare and download. Each ZIP’s clock starts when that run finishes.
DIFFERENT TOOLS. DIFFERENT OUTPUTS.
Pick the structure you need.
Docling
Structured documents with HTML and DocTags alongside Markdown.
Planned version 2.124.0 · Up to 500 pages
Mistral OCR
Hosted OCR with page-level text and structured blocks.
Planned version 4.1 · Up to 100 pages
Output quality and structure vary by processor and document. The planned service sends your PDF to your selected processors and makes their returned files available. A technical failure triggers a refund of that engine’s share; completed runs have no automatic quality refund.