OCRRouter

PaddleOCR-VL vs Nougat

Pick PaddleOCR-VL for multilingual scans, mixed business and report pages, and any commercial product. Pick Nougat only for English, math-heavy scientific papers you want as LaTeX, and only if your use is non-commercial. The deciding factor is licence and scope: PaddleOCR-VL is Apache-2.0 and covers 109 languages, while Nougat's weights are CC-BY-NC and it fails on non-Latin scripts.

PaddleOCR-VLpaddle/paddleocr-vlNougatmeta/nougat
TypeOpen sourceOpen source
Price6 credits / page2 credits / page
Per 1,000 pages$6.00$2.00
Typical speed1.3 s / page3.0 s / page
Best forCJK documentsarXiv-style papers
InputsPDF, imagesPDF, images
Languages109 languagesMainly English
Max pages per request500200
Tables✓ Yes✓ Yes
HandwritingNoNo
Math✓ Yes✓ Yes
Bounding boxes✓ YesNo
Confidence✓ YesNo
FormsNoNo
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 PaddleOCR-VL if

  • It is Apache-2.0, so you can use it in a commercial product.
  • It supports 109 languages, including Cyrillic, Arabic, Devanagari and Thai.
  • It parses text, tables, formulas and charts into Markdown and JSON, and scores 96.33 on OmniDocBench v1.6 (author-reported).

Choose Nougat if

  • It is built for English arXiv-style papers and was trained on arXiv and PubMed Central documents.
  • It outputs LaTeX for equations and tables in Mathpix-compatible .mmd Markdown, which suits math-heavy PDFs.
  • It has small and base checkpoints and a simple CLI and API, for non-commercial self-hosting.

Key differences

PaddleOCR-VLNougat
LicenceApache-2.0Code MIT; weights CC-BY-NC (no commercial use)
Deployment and priceSelf-host via PaddleOCR, vLLM or transformers; free, you pay for computeSelf-host Python package; free for non-commercial use, you pay for compute
Output formatsMarkdown, JSONMarkdown (.mmd), LaTeX
Languages109 languages, including Cyrillic, Arabic, Devanagari and ThaiWorks best on English; README says Chinese, Russian and Japanese do not work
Target documentsScanned documents, academic papers, tables and charts in reportsEnglish scientific papers; not suited to business documents, forms and invoices
BenchmarksOmniDocBench v1.6: 96.33 (authors); olmOCR-Bench: 80.0 (independent)Not published
Known failureVersion 1.6 has a Cyrillic regression on born-digital PDFs, hangs on blurry images, and ships use_cache false (about 11x slower decoding)Repetition loops on 1.5% of test pages, more on out-of-domain documents; looping pages can be dropped as MISSING_PAGE

Sources

What each one is

PaddleOCR-VL

Open-source document model from Baidu. Excellent on Chinese, Japanese and Korean.

PaddleOCR-VL details

Nougat

Meta's model that converts academic PDFs to Markdown with LaTeX math.

Nougat 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": "paddle/paddleocr-vl","fallbacks": ["meta/nougat"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

PaddleOCR-VL alternativesNougat alternativesBest OCR models in 2026