Jina OCR vs Donut
Pick Jina OCR for turning full PDF pages into Markdown, HTML or LaTeX, as long as the use is non-commercial. Pick Donut for extracting fields from a fixed type of document such as receipts or forms, when you can fine-tune it and need an MIT licence. The main split is licence and readiness: Jina OCR works on pages out of the box but blocks commercial use, while Donut is free for any use but must be trained first.
Jina OCRjina/jina-ocr | Donutnaver/donut | |
|---|---|---|
| Type | Open source | Open source |
| Price | 1 credit / page | 1 credit / page |
| Per 1,000 pages | $1.00 | $1.00 |
| Typical speed | 1.5 s / page | 1.0 s / page |
| Best for | Clean Markdown at low cost | Receipt and form fields after fine-tuning |
| Inputs | PDF, images | images |
| Languages | About 100 languages | English, Chinese, Japanese, Korean |
| Max pages per request | 200 | 100 |
| Tables | ✓ Yes | No |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| Bounding boxes | No | No |
| Confidence | No | No |
| Forms | No | ✓ Yes |
| Multilingual | ✓ Yes | 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 Jina OCR if
- You need full-page PDF to Markdown, HTML or LaTeX output, which Donut does not target.
- You want a published score: 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench, both reported by the authors.
- You want a hosted API option as well as self-hosting, with only about 570M active parameters and 2.57 pages per second on an A100.
Choose Donut if
- You are building a commercial product and need an MIT licence with no extra deal.
- You process a narrow, fixed-layout document type, such as receipts or forms, and can fine-tune on it.
- You want JSON output straight from the image with no separate OCR step.
Key differences
| Jina OCR | Donut | |
|---|---|---|
| Licence | CC-BY-NC-4.0, commercial use needs a license from Jina/Elastic | MIT |
| How you run it | Self-host or Jina's hosted API | Self-host only |
| Ready to use | Open-weight page-to-Markdown model | Base checkpoint is pretrain-only and must be fine-tuned first |
| Output formats | Markdown, HTML, LaTeX | JSON, plain text |
| Benchmarks | 91.14 OmniDocBench v1.6, 83.4 olmOCR-Bench (authors) | Not published |
| Model size | 3B MoE, about 570M active | Not published |
| Known failure | Hosted API returns HTTP 503 on cold start; retry after 30 to 60 seconds | Misses tiny text in large images due to input resolution limits |
Sources
What each one is
Jina OCR
Compact page-to-Markdown model from Jina AI, built on the DeepSeek-OCR encoder.
Jina OCR 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": "jina/jina-ocr","fallbacks": ["naver/donut"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Jina OCR alternativesDonut alternativesBest OCR models in 2026