Donut vs Unstructured
Pick Unstructured if you need to ingest mixed file types such as PDFs, Word, PowerPoint, HTML and emails for RAG, or want a hosted API with 10,000 free pages. Pick Donut if you have a narrow, fixed-layout document type such as receipts or forms and can fine-tune a model yourself. The main split is that Donut is a base model you must train, while Unstructured is a ready-to-run partitioning pipeline.
Donutnaver/donut | Unstructuredunstructured/unstructured | |
|---|---|---|
| Type | Open source | OCR API |
| Price | 1 credit / page | 20 credits / page |
| Per 1,000 pages | $1.00 | $20.00 |
| Typical speed | 1.0 s / page | 3.0 s / page |
| Best for | Receipt and form fields after fine-tuning | Mixed file types into RAG |
| Inputs | images | PDF, images, DOCX |
| Languages | English, Chinese, Japanese, Korean | Multilingual |
| Max pages per request | 100 | 1,000 |
| Tables | No | ✓ Yes |
| Handwriting | No | No |
| Math | No | No |
| Bounding boxes | No | ✓ Yes |
| Confidence | No | No |
| Forms | ✓ Yes | No |
| Multilingual | No | ✓ Yes |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose Donut if
- You can fine-tune on a narrow, fixed-layout set such as receipts or forms, and a practitioner on HN reports about 200 documents can reach human-level accuracy on such data.
- You want one model that maps the document image straight to structured JSON with no external OCR step.
- You want a permissive MIT licence and loading through Hugging Face Transformers, with all compute under your control.
Choose Unstructured if
- You need to ingest PDFs, HTML, Word, PowerPoint, emails and images through one library.
- You want a hosted API: the first 10,000 pages are free, then $0.015 per page.
- You work mostly with born-digital PDFs, where the fast strategy applies, or you want to trade speed for quality with the Auto, Fast, High-Res and VLM strategies.
Key differences
| Donut | Unstructured | |
|---|---|---|
| What it is | OCR-free Swin encoder plus BART decoder transformer from NAVER Clova (ECCV 2022). Base checkpoint is pretrain-only. | Document ETL library and platform that partitions files into typed elements. |
| Ready to use | No. It must be fine-tuned for parsing, classification or VQA before use. | Yes. Choose a strategy (Auto, Fast, High-Res or VLM) and run it. |
| Licence | MIT | Apache-2.0 for the open-source library; proprietary for the hosted platform. |
| How you run it | Self-host only. | Self-host the library or call the hosted API. |
| Price | Free; you pay for compute. | Hosted: first 10,000 pages free, then $0.015 per page. Business is custom. Open-source library is free. |
| Output formats | JSON, plain text. | JSON, HTML, plain text, bounding boxes. |
| Known failure | Misses tiny text in large images because of input resolution limits. Transformers v5 dropped image-to-text pipeline support, so you need direct loading or v4. | Open-source tables are weak: table cell content scores 0.426 versus 0.866 hosted (vendor's own benchmark). hi_res tables can cut off at the edges and drop first letters, and table OCR ignores language settings. |
Sources
- donut-base model card
- Donut ECCV paper
- Unstructured pricing
- Unstructured GitHub README
- Unstructured issue #3262
- Hacker News thread
What each one is
Donut
OCR-free document understanding model. Fine-tuned for receipts, forms and document Q&A.
Donut detailsUnstructured
Document ETL for RAG. Partitions many file types into typed elements; open-source library too.
Unstructured detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "naver/donut","fallbacks": ["unstructured/unstructured"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Donut alternativesUnstructured alternativesBest OCR models in 2026