OCRRouter

Hunyuan OCR vs Nougat

Pick Nougat for English scientific papers where you need LaTeX math and tables, as long as the project is non-commercial. Pick Hunyuan OCR for scene text, translation, low-resource scripts or a general document parser, but check that the Tencent licence covers your region.

Hunyuan OCRtencent/hunyuan-ocrNougatmeta/nougat
TypeOpen sourceOpen source
Price2 credits / page2 credits / page
Per 1,000 pages$2.00$2.00
Typical speed1.5 s / page3.0 s / page
Best forChinese documents and photosarXiv-style papers
InputsPDF, imagesPDF, images
Languages100+ languagesMainly English
Max pages per request500200
Tables✓ Yes✓ Yes
HandwritingNoNo
Math✓ Yes✓ Yes
Bounding boxes✓ YesNo
ConfidenceNoNo
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 Hunyuan OCR if

  • It handles photo and scene text spotting, text-image translation, information extraction and document parsing in one ~1B model.
  • It supports low-resource and ancient scripts, and exposes 12 task types via the vLLM client.
  • It can run on a laptop or consumer GPU through the llama.cpp GGUF path, and it returns Markdown, HTML, LaTeX, JSON and bounding boxes.

Choose Nougat if

  • It is built for academic papers, trained on arXiv and PubMed Central, and outputs LaTeX for equations and tables in Mathpix-compatible Markdown.
  • It has small and base checkpoints with a simple CLI and API, so setup for papers is short.
  • The code is MIT, which makes it easy to read and modify, though the weights are CC-BY-NC.

Key differences

Hunyuan OCRNougat
LicenceTencent Hunyuan Community License. Excludes the EU, UK and South Korea. A separate licence is needed above 100M monthly active users.Code is MIT. Weights are CC-BY-NC, so no commercial use.
Best document typePhotos, scene text, text-image translation, historical and low-resource scripts.English arXiv-style scientific papers and math-heavy PDFs.
How you run itSelf-hosted via vLLM, Transformers or llama.cpp. CUDA 13 recommended for the unified vLLM environment.Self-hosted Python package with a CLI and API.
Output formatsMarkdown, HTML, LaTeX, JSON, plain text, bounding boxes.Markdown (.mmd) and LaTeX.
LanguagesIncludes low-resource and ancient-script OCR.Works best on English. The README says Chinese, Russian and Japanese do not work.
Known failureArchived 1.0 note said Transformers and vLLM outputs differed in accuracy. ParseBench-2 overall is 67.11.Repetition loops on 1.5% of test pages in the paper, more on out-of-domain documents. Pages can be dropped as MISSING_PAGE.
PriceFree, self-hosted. You pay for compute.Free for non-commercial use, self-hosted. You pay for compute.

Sources

What each one is

Hunyuan OCR

Small end-to-end OCR model from Tencent covering parsing, extraction and photo translation.

Hunyuan OCR 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": "tencent/hunyuan-ocr","fallbacks": ["meta/nougat"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

Hunyuan OCR alternativesNougat alternativesBest OCR models in 2026