OCRRouter

SmolDocling vs Unstructured

Pick Unstructured if you ingest mixed files (PDF, HTML, Word, PowerPoint, email, images) into a RAG pipeline, or want a hosted API. Pick SmolDocling only for experiments or low-budget English page conversion on your own hardware. SmolDocling is a preview release, and its authors point users to Granite-Docling for support.

SmolDoclinghuggingface/smoldoclingUnstructuredunstructured/unstructured
TypeOpen sourceOCR API
Price1 credit / page20 credits / page
Per 1,000 pages$1.00$20.00
Typical speed1.0 s / page3.0 s / page
Best forVery cheap page conversionMixed file types into RAG
InputsPDF, imagesPDF, images, DOCX
LanguagesMainly EnglishMultilingual
Max pages per request5001,000
Tables✓ Yes✓ Yes
HandwritingNoNo
Math✓ YesNo
Bounding boxes✓ Yes✓ Yes
ConfidenceNoNo
FormsNoNo
MultilingualNo✓ Yes
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 SmolDocling if

  • It is a 256M-parameter model that runs on modest hardware, at about 0.35 seconds per page on an A100 with vLLM.
  • One model handles text, tables, code, formulas and charts, and returns location tags and bounding boxes.
  • It is free under an Apache-2.0 licence per the model summary, and you only pay for compute.

Choose Unstructured if

  • It reads PDFs, HTML, Word, PowerPoint, emails and images in one library.
  • You can start with 10,000 free hosted pages, then pay $0.015 per page, or self-host the Apache-2.0 library.
  • Its Auto, Fast, High-Res and VLM strategies let you trade speed for quality, and Fast suits born-digital PDFs.

Key differences

SmolDoclingUnstructured
DeploymentSelf-host onlySelf-host the open-source library or use the hosted API
LicenceApache-2.0 per model summary (metadata tag says CDLA-Permissive-2.0)Apache-2.0 for the library; proprietary for the hosted platform
PriceFree, you pay for computeHosted: first 10,000 pages free, then $0.015 per page; library is free
Input scopePage images converted to DocTagsPDFs, HTML, Word, PowerPoint, emails and images
Output formatsDocTags, Markdown, HTML, JSON, LaTeX, bounding boxesJSON, HTML, plain text, bounding boxes
Table qualityTEDS structure 0.82 on FinTabNet at 150dpi (authors' figure)Table cell content 0.426 for the open-source library vs 0.866 hosted (vendor's figure)
Known failureOutput can loop on some images and become very slow (Docling issue #1515); English onlyhi_res tables can cut off at the edges and drop first letters (issue #3262); table OCR ignores language settings (issue #1694)

Sources

What each one is

SmolDocling

Tiny 256M vision model from Hugging Face and IBM that outputs DocTags. Predecessor of Granite Docling.

SmolDocling details

Unstructured

Document ETL for RAG. Partitions many file types into typed elements; open-source library too.

Unstructured 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": "huggingface/smoldocling","fallbacks": ["unstructured/unstructured"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

SmolDocling alternativesUnstructured alternativesBest OCR models in 2026