parsebench

PDF → Markdown, measured

parsebench

Every parser’s README claims it handles tables well. Race 2–4 of them against your own PDF and read the Markdown side by side — with time, cost, OCR and licence attached to every result.

Run your own PDFSee the comparison
parsers compared
3
per run
2–4
page limit
100
fastest in the demo
pdf-inspector

The same table, three parsers

Two of these kept the table. One turned it into a row of numbers.

Real output, captured from tables.pdf — one page, one financial table — three regions by three quarters. Nothing below is edited or illustrative.

pdf-inspector

13ms

Financial Summary

pdf-inspector output: the source table, preserved
RegionQ1Q2Q3
North120145160
South98102115
East210198225

Table preserved

LiteParse

304ms

Financial Summary

LiteParse output: the source table, preserved
RegionQ1Q2Q3
North120145160
South98102115
East210198225

Table preserved

OpenDataLoader PDF

405ms

Financial Summary

Region Q1 Q2 Q3 North 120 145 160 South 98 102 115 East 210 198 225

Table flattened

Once a table is flattened, the numbers have lost their headers. Ask a RAG pipeline built on that third output “What was North's Q2?” and there is nothing left in the text that ties 145 to either North or Q2. The retrieval was never the problem; the parse was.

Your turn

Run it on your own PDF

The demo above is one small table. Your documents are the ones that matter — the invoice with merged cells, the scan with no text layer, the 80-page report. Drop one in and watch the parsers race it.

Before you upload

  • The PDF you upload is deleted after 24 hours.
  • The converted Markdown is kept indefinitely so the run link keeps working — which means the document’s text is retained after the PDF itself is gone.
  • The run link is unlisted but not private: anyone who has the URL can read the results.

Drop a PDF here, or .

PDF · up to 100 pages · 25 MB

The field

What each parser can actually do

Read straight off each adapter’s declared capabilities, with the licence verified against the package it ships. A parser with no OCR engine cannot read a scan — parsebench says so before you wait for the run, not after.

Parser capabilities: OCR support, page markers, licence, cost and observed speed
ParserOCRPage markersLicenceCostDemo time
pdf-inspectorpdf-inspectorno OCRyesMITfree13ms
LiteParseliteparsebuilt-in OCRyesApache-2.0free304ms
OpenDataLoader PDFopendataloaderbuilt-in OCRyesApache-2.0free405ms

Demo time is one run of one small table on one machine. It sets the order of magnitude, nothing more — the run you start above is measured on your own document.

How it works

Four steps, about ninety seconds

  1. 01

    Upload

    A PDF, up to 100 pages. It never leaves the run it belongs to.

  2. 02

    Classify

    parsebench reads the text layer first and tells you whether the document is text-based, scanned, or mixed.

  3. 03

    Race

    Pick 2–4 parsers. They run in parallel and each pane fills the moment its result lands.

  4. 04

    Compare

    Rendered Markdown side by side, with time, cost, OCR and licence attached to every column.