PaddleOCR-VL vs Donut
Pick PaddleOCR-VL when you need full-page parsing of scans, papers, tables and charts into Markdown or JSON, with no training step. Pick Donut when you have a narrow, fixed-layout document type such as receipts or forms and can fine-tune it. The main split is that PaddleOCR-VL is a ready document parser, while Donut is a pretrain-only base model that must be trained first.
PaddleOCR-VLpaddle/paddleocr-vl | Donutnaver/donut | |
|---|---|---|
| Type | Open source | Open source |
| Price | 6 credits / page | 1 credit / page |
| Per 1,000 pages | $6.00 | $1.00 |
| Typical speed | 1.3 s / page | 1.0 s / page |
| Best for | CJK documents | Receipt and form fields after fine-tuning |
| Inputs | PDF, images | images |
| Languages | 109 languages | English, Chinese, Japanese, Korean |
| Max pages per request | 500 | 100 |
| Tables | ✓ Yes | No |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | ✓ Yes | 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 PaddleOCR-VL if
- It parses whole pages into Markdown and JSON, covering text, tables, formulas and charts.
- It supports 109 languages, including Cyrillic, Arabic, Devanagari and Thai scripts, in one model.
- At 0.9B parameters it runs on much smaller GPUs than 7B document VLMs, and the authors report 96.33 on OmniDocBench v1.6 for version 1.6.
Choose Donut if
- It maps document images straight to structured JSON with no external OCR step, which suits receipt and form field extraction after fine-tuning.
- It is MIT licensed and available in Hugging Face Transformers.
- A practitioner on Hacker News reports that about 200 documents of fine-tuning can reach human-level accuracy on narrow data.
Key differences
| PaddleOCR-VL | Donut | |
|---|---|---|
| Licence | Apache-2.0 | MIT |
| Ready to use | Yes, it parses documents as released | No, the base checkpoint is pretrain-only and must be fine-tuned first |
| Output formats | Markdown, JSON | JSON, plain text |
| Benchmarks | OmniDocBench v1.6 overall 96.33 (authors, version 1.6); olmOCR-Bench overall 80.0 (independent) | Not published |
| Best document types | Multilingual scans, academic papers with formulas, tables and charts in reports | Receipts, forms, classification and other fixed-layout documents |
| Hardware | NVIDIA GPU recommended; CPU possible but slow; no minimum VRAM stated | Not published |
| Known failure | Version 1.6 has a Cyrillic regression on born-digital PDFs, hangs on blurry images, and ships with use_cache false, which makes decoding about 11x slower | Misses tiny text in large images because of input resolution limits; Transformers v5 dropped image-to-text pipeline support |
Sources
- Hugging Face model card
- PaddleOCR-VL-1.6 model card
- olmOCR-Bench table
- donut-base model card
- Donut ECCV paper
- Hacker News thread
What each one is
PaddleOCR-VL
Open-source document model from Baidu. Excellent on Chinese, Japanese and Korean.
PaddleOCR-VL 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": "paddle/paddleocr-vl","fallbacks": ["naver/donut"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
PaddleOCR-VL alternativesDonut alternativesBest OCR models in 2026