extend

Drag 'n' drop here, or click to select files

Start a batch run with multiple documents

Configuration Builder

Set output format, chunking, and block options

Engine & Format

Engine

Engine determines the models used in the pipeline. Pin a specific parser base version for reproducibility. Learn more here.

Light

Much cheaper and often faster parsing. Best for high-volume ingestion and simpler documents.

0.5 CR/page

Performance

Max-accuracy parsing. Recommended default for complex documents, large tables, multilingual, and forms.

2 CR/page

Engine version

Select the version of the engine. Different versions can control things like our layout model, OCR models, available advanced options, and more. Think of it like a model version.

Target Format

Select the output format for parsed content

Parses the document into a markdown target format, including logical markdown elements and hierarchy. Generally recommended for most use cases.

Figures

Enable figure parsing

Enable image clipping

Enable advanced chart data extraction

Text

Enable advanced signature detection

Enable agentic OCR mode

Tables

Enable agentic table mode

Enable table header continuation

Return table cell blocks as children

Table target format

Key-Value

Empty field formatting

Requires Performance 2.0.0+

Barcodes

Enable barcode/QR code image clipping

Requires Performance 2.0.0+

Enable barcode/QR code reading and decoding

Requires Performance 2.0.0+

Formulas

Enable math parsing (LaTeX extraction)

Requires Performance 2.0.0+

Change tracking (strike through, insertions)

Requires Performance 2.0.0+