Choose a parser by where documents may be processed and which output you need. Text extraction, OCR and layout recovery are capabilities that can be combined within one tool; they are not mutually exclusive product categories.

Current capability map

Checked 10 September 2026 against official documentation. These are capability descriptions, not a quality or price ranking.

Tool	Documented option	Integration consequence
PyMuPDF	Integrated OCR via separately installed Tesseract	Use `Page.get_textpage_ocr` when a text layer is absent; OCR requires setup
PyMuPDF4LLM	Markdown extraction with OCR support	A local Markdown workflow is available
pdfplumber	Table strategies based on lines, text or explicit edges	Tune extraction to the document; visible gridlines are not mandatory
Camelot	pdfium is the default image backend since 1.0.0	Ghostscript is optional, not a universal prerequisite
Tesseract	Text recognition with selected trained language data	Install the languages your workflow needs
Azure Document Intelligence	Layout output with `outputContentFormat=markdown`	Markdown and structured HTML tables are available in documented configurations
LlamaParse	Markdown, text and JSON parsing output	Choose the format your downstream stack needs
Unstructured	Typed elements with configurable PDF strategies	Keep element metadata when your workflow needs it
pdfToMarkdown	Hosted PDF conversion to complete Markdown	Validate success before consuming content; retain account request identities

Sources: PyMuPDF OCR, PyMuPDF4LLM, pdfplumber, Camelot installation, Tesseract, Azure Markdown, LlamaParse, Unstructured.

Run a useful evaluation

Select text-native PDFs, scans and your most difficult tables. Compare missing text, reading order, merged cells and downstream parsing effort. Measure total elapsed time on the client. Use current pricing when estimating cost.

Current and historical pdfToMarkdown evidence

Mistral OCR 4.1 is the current processor. Inspect the dated invoice capture for its source and complete output. The July 2026 CJK benchmark and PaddleOCR guide remain historical evidence for the earlier implementation; their scores are not current Mistral measurements.

Convert a document

Use the tested Python, JavaScript or curl workflow. Choose one input: a public PDF URL, Base64 bytes, or a raw PDF upload. Raw uploads avoid Base64 expansion; all three send the document to an external processor.

The public demo converts page 1 with a watermark, at 3 requests per minute per IP. New accounts receive 20 trial pages once. Further account conversions use paid credits: monthly pages expire, top-ups do not. The API reference owns authentication, limits, replay and retention.