OCRRouter

Marker vs Docling

Pick Marker to turn PDFs of books and papers with equations into Markdown on your own hardware, if the licence terms fit your company. Pick Docling to ingest born-digital PDFs and Office files under an MIT licence into a structured JSON document for RAG. The main split is scope and licence: Marker is a PDF-focused converter with a restricted weight licence, while Docling covers many formats and its own code is MIT.

Markerdatalab/markerDoclingibm/docling
TypeOCR APIOpen source
Price14 credits / page2 credits / page
Per 1,000 pages$14.00$2.00
Typical speed2.5 s / page2.2 s / page
Best forBooks and research papersStructured export for RAG
InputsPDF, images, DOCXPDF, images, DOCX
Languages90+ languagesLatin scripts, CJK
Max pages per request1,000500
Tables✓ Yes✓ Yes
HandwritingNoNo
Math✓ Yes✓ Yes
Bounding boxes✓ Yes✓ Yes
ConfidenceNoNo
FormsNoNo
Multilingual✓ YesNo
On OCRRouterAvailableAvailable

Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.

Which to choose

Choose Marker if

  • Datalab reports 83.7 on olmOCR-bench in vllm mode, against 50.3 for Docling, though these are Datalab's own numbers.
  • It suits books and papers with equations, and it outputs LaTeX and RAG chunks.
  • Fast mode runs on CPU at 23.6 pages per second with about 18 MB of ONNX weights, and a hosted API is available with about 5,000 free pages per month.

Choose Docling if

  • The MIT licence has no revenue cap, so a large company can use it commercially without a paid licence.
  • One API parses PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, emails and XBRL.
  • Lossless JSON export keeps layout, reading order and table structure, and it plugs into LangChain and LlamaIndex.

Key differences

MarkerDocling
LicenceCode Apache-2.0. Weights under a modified AI Pubs Open Rail-M licence, free for research, personal use and startups under $5M.MIT. Individual models have their own licences.
How you run itLocal on GPU, CPU or Apple Silicon. Datalab also sells a hosted API.Local Python library on macOS, Linux and Windows, x86_64 and arm64. CPU works, GPU speeds up model stages.
Input formatsPDF to Markdown converter.PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, emails and XBRL.
Output formatsMarkdown, HTML, JSON, chunks, LaTeX, images.Markdown, HTML, JSON, DocTags, plain text.
BenchmarkolmOCR-bench 83.7 in vllm mode, 66.6 in CPU fast mode. All figures are from Datalab.Not published
Scanned pagesvllm mode needs an NVIDIA GPU. Fast mode on CPU scores lower.Accuracy depends on the OCR engine you plug in: EasyOCR, Tesseract, RapidOCR or a VLM such as the 258M GraniteDocling.
Known failureWeight licence blocks free commercial use above $5M without a paid licence.Version 2.143.1 drops words and numbers with full-page Tesseract OCR (issue #5254). Reading order can merge neighbouring columns (issue #5132).

Sources

What each one is

Marker

Hosted version of the open-source Marker and Surya models. Strong on books and papers.

Marker details

Docling

IBM's open-source document converter with a structured document model.

Docling details

Try both on your own file

Run the file once with each model and compare the output.

PDF, PNG, JPG, WebP, HEIC. Up to 20 MB each, 10 files.No file?

Use one, fall back to the other

{"model": "datalab/marker","fallbacks": ["ibm/docling"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

Marker alternativesDocling alternativesBest OCR models in 2026