OCRRouter

dots.ocr vs Donut

Pick dots.ocr to turn full pages (PDFs, reports, papers) into Markdown or HTML with layout boxes, tables and formulas, with no training. Pick Donut only for a narrow, fixed-layout document type such as receipts or forms, where you can fine-tune it on your own samples and need JSON fields back. The main split is that dots.ocr works on whole pages out of the box, while Donut is a base model you must train first.

dots.ocrrednote/dots-ocrDonutnaver/donut
TypeOpen sourceOpen source
Price2 credits / page1 credit / page
Per 1,000 pages$2.00$1.00
Typical speed1.5 s / page1.0 s / page
Best forMultilingual document parsingReceipt and form fields after fine-tuning
InputsPDF, imagesimages
Languages100+ languagesEnglish, Chinese, Japanese, Korean
Max pages per request500100
Tables✓ YesNo
HandwritingNoNo
Math✓ YesNo
Bounding boxes✓ YesNo
ConfidenceNoNo
FormsNo✓ Yes
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 dots.ocr if

  • It returns layout boxes, categories and text in one pass, with tables as HTML and formulas as LaTeX.
  • It has published scores: 79.1 on olmOCR-bench for the original model and 83.9 for dots.mocr (v1.5).
  • It handles multilingual PDFs, and the dots.mocr-svg variant converts charts and diagrams to SVG.

Choose Donut if

  • It maps a document image straight to JSON with no separate OCR step, which suits receipt and form field extraction after fine-tuning.
  • A practitioner on Hacker News reports that about 200 documents of fine-tuning can reach human-level accuracy on narrow data.
  • It is MIT licensed and available in Hugging Face Transformers, so you can adapt it to a fixed layout and own the resulting model.

Key differences

dots.ocrDonut
Ready to useParses documents out of the boxBase checkpoint is pretrain-only and must be fine-tuned first
Output formatsJSON, Markdown, HTML, LaTeX, bounding boxes, SVGJSON, plain text
Model size3B parameters (1.7B LLM backbone)Not published
BenchmarksolmOCR-bench 79.1 (dots.ocr), 83.9 (dots.mocr); OmniDocBench v1.5 text edit distance 0.031 (dots.mocr, lower is better)Not published
Licence and costMIT code, with a separate dots.ocr licence agreement file in the repository; free, you pay for computeMIT; free, you pay for compute
Known failureRuns of ellipses or underscores can make output repeat endlessly; complex tables and formulas remain difficult; one table above 1000 grid cells raised an error and dropped the whole page in a Docling issueMisses tiny text in large images because of input resolution limits; one HN user found results on personal documents did not match the promised scores
Setup notesSingle CUDA GPU with vLLM is the documented path; CPU inference guide is linkedTransformers v5 dropped image-to-text pipeline support; load the model directly or use v4

Sources

What each one is

dots.ocr

Single vision model from rednote that does layout detection and content recognition together.

dots.ocr details

Donut

OCR-free document understanding model. Fine-tuned for receipts, forms and document Q&A.

Donut 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": "rednote/dots-ocr","fallbacks": ["naver/donut"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

dots.ocr alternativesDonut alternativesBest OCR models in 2026