OCRRouter

Google Document AI vs SmolDocling

Pick Google Document AI if you run on Google Cloud and need a supported, hosted API for bulk OCR or RAG chunking. Pick SmolDocling only for experiments with a tiny self-hosted model on English pages. SmolDocling is a preview release that its authors have replaced with Granite-Docling, so it is not a production choice.

Google Document AIgoogle/document-aiSmolDoclinghuggingface/smoldocling
TypeOCR APIOpen source
Price10 credits / page1 credit / page
Per 1,000 pages$10.00$1.00
Typical speed1.4 s / page1.0 s / page
Best forMultilingual scansVery cheap page conversion
InputsPDF, imagesPDF, images
Languages200+ languagesMainly English
Max pages per request500500
Tables✓ Yes✓ Yes
Handwriting✓ YesNo
MathNo✓ Yes
Bounding boxes✓ Yes✓ Yes
Confidence✓ YesNo
Forms✓ YesNo
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 Google Document AI if

  • Enterprise Document OCR costs $1.50 per 1,000 pages, dropping to $0.60 above 5M pages a month, with the first 1,000 pages free.
  • Layout Parser includes chunking for RAG at $10 per 1,000 pages.
  • Batch requests take up to 500 pages, so large OCR jobs fit in one call.

Choose SmolDocling if

  • It is free to run, and you only pay for your own compute.
  • At 256M parameters it fits modest hardware, and it runs at about 0.35 seconds per page on an A100 with vLLM.
  • One model returns text, tables, code, formulas and charts with location tags, in formats such as Markdown, HTML, JSON and LaTeX.

Key differences

Google Document AISmolDocling
How you run itHosted API on Google CloudSelf-hosted, 256M parameters
LicenceProprietaryApache-2.0 per the model summary (the metadata tag says CDLA-Permissive-2.0)
Price$1.50 per 1,000 pages for Enterprise OCR, $0.60 above 5M a month; Layout Parser $10; Form Parser $30Free, you pay for compute
Output formatsJSON, plain text, bounding boxes, chunksDocTags, Markdown, HTML, JSON, LaTeX, bounding boxes
Table accuracy64.6% on RD-TableBench (independent, behind Azure and Textract)TEDS structure 0.82 on FinTabNet at 150dpi (reported by the authors, not comparable to RD-TableBench)
Page and size limitsOnline requests limited to 15 pages (30 in imageless mode); batch up to 500 pages; Form Parser batch up to 100 pagesNot published
Languages and known failuresNot publishedEnglish only; endless token repetition and missing closing tags are typical defects, and Docling issue #1515 reports output loops that make generation very slow

Sources

What each one is

Google Document AI

Google Cloud OCR with layout parsing and specialised processors.

Google Document AI details

SmolDocling

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

SmolDocling 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": "google/document-ai","fallbacks": ["huggingface/smoldocling"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

Google Document AI alternativesSmolDocling alternativesBest OCR models in 2026