Amazon Textract vs Donut
Pick Amazon Textract if you run on AWS and need English forms, invoices or tables from large scanned PDFs through a managed API. Pick Donut if you have a narrow, fixed-layout document type such as receipts, can fine-tune a model, and want an MIT-licensed model you host yourself. The main split is a ready-to-call paid service versus a free base model that does nothing useful until you train it.
Amazon Textractaws/textract | Donutnaver/donut | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 25 credits / page | 1 credit / page |
| Per 1,000 pages | $25.00 | $1.00 |
| Typical speed | 1.2 s / page | 1.0 s / page |
| Best for | Forms and key-value extraction | Receipt and form fields after fine-tuning |
| Inputs | PDF, images | images |
| Languages | 6 languages | English, Chinese, Japanese, Korean |
| Max pages per request | 3,000 | 100 |
| Tables | ✓ Yes | No |
| Handwriting | ✓ Yes | No |
| Math | No | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | ✓ Yes | No |
| Forms | ✓ Yes | ✓ Yes |
| Multilingual | No | No |
| 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 Amazon Textract if
- It works through the AWS API with no training, and basic text detection costs $1.50 per 1,000 pages.
- The async API takes PDFs up to 500 MB and 3,000 pages, which suits large scanned batches.
- It extracts tables, forms and queries, and returns word and line geometry with confidence scores.
Choose Donut if
- It is MIT licensed and free, so you pay only for your own compute and keep documents on your own infrastructure.
- It maps document images straight to structured JSON with no separate OCR step, which suits receipt and form field extraction after fine-tuning.
- It can be fine-tuned for classification or parsing on fixed-layout documents, and one HN practitioner reports about 200 documents can reach human-level accuracy on narrow data.
Key differences
| Amazon Textract | Donut | |
|---|---|---|
| How you run it | Hosted AWS API | Self-hosted model |
| Licence | Proprietary | MIT |
| Price | US West (Oregon), first 1M pages/month: text $0.0015/page, Tables $0.015, Forms $0.05, Queries $0.015 | Free, you pay for compute |
| Ready to use | Yes, call the API | No, the base checkpoint is pretrain-only and must be fine-tuned |
| Languages | Text detection: English, French, German, Italian, Portuguese, Spanish. Handwriting and Queries: English only | Not published |
| Table benchmark | 80.9% on RD-TableBench (independent) | Not published |
| Known failure | Vertical Japanese and Chinese text is not supported; the sync API takes one page | Misses tiny text in large images because of input resolution limits; one HN user's results on personal documents did not match the promised scores |
Sources
- Textract pricing
- Textract quotas
- Reducto RD-TableBench
- donut-base model card
- Donut ECCV paper
- Hacker News thread
What each one is
Amazon Textract
AWS OCR with form key-value pairs, tables and per-word confidence scores.
Amazon Textract detailsDonut
OCR-free document understanding model. Fine-tuned for receipts, forms and document Q&A.
Donut 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": "aws/textract","fallbacks": ["naver/donut"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Amazon Textract alternativesDonut alternativesBest OCR models in 2026