Hunyuan OCR vs Donut
Pick Hunyuan OCR for photo and scene text, text-image translation, and low-resource or historical scripts, and when you want a model that works without training. 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 Hunyuan OCR is a ready ~1B model with many tasks and outputs, while Donut is an MIT-licensed base checkpoint that must be fine-tuned before use.
Hunyuan OCRtencent/hunyuan-ocr | Donutnaver/donut | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 1 credit / page |
| Per 1,000 pages | $2.00 | $1.00 |
| Typical speed | 1.5 s / page | 1.0 s / page |
| Best for | Chinese documents and photos | Receipt and form fields after fine-tuning |
| Inputs | PDF, images | images |
| Languages | 100+ 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 | 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 Hunyuan OCR if
- It handles document parsing, text spotting, information extraction and image translation in one ~1B model, with no fine-tuning step.
- It outputs Markdown, HTML, LaTeX, JSON, plain text and bounding boxes, so it can convert full pages.
- It covers low-resource and ancient-script OCR, supports images up to 4K, and runs on laptops via llama.cpp GGUF.
Choose Donut if
- It uses the MIT licence, while Hunyuan OCR's licence excludes the EU, UK and South Korea and needs a separate licence above 100M monthly active users.
- It maps document images straight to structured JSON with no external OCR step, which suits receipt and form field extraction after fine-tuning.
- A practitioner on Hacker News reports that about 200 documents of fine-tuning can reach human-level accuracy on narrow data.
Key differences
| Hunyuan OCR | Donut | |
|---|---|---|
| Licence | Tencent Hunyuan Community License; excludes EU, UK and South Korea; separate licence above 100M monthly active users | MIT |
| Ready to use | Covers parsing, spotting, extraction and translation in one model | Base checkpoint is pretrain-only and must be fine-tuned first |
| Output formats | Markdown, HTML, LaTeX, JSON, plain text, bounding boxes | JSON, plain text |
| Model size and hardware | ~1B parameters; CUDA 13 GPU recommended for the vLLM setup, llama.cpp GGUF for CPU, consumer GPU or laptop | Not published |
| Benchmarks | ParseBench-2 overall 67.11 (v1.5); ParseBench mean 63.86 | Not published |
| Best document types | Photo and scene text, text-image translation, historical scripts | Receipts, forms, classification, fixed-layout documents |
| Known failure | Archived 1.0 note said Transformers and vLLM outputs differed in accuracy | Misses tiny text in large images due to input resolution limits; Transformers v5 dropped image-to-text pipeline support |
Sources
- HunyuanOCR model card
- Hunyuan Community License
- donut-base model card
- Donut ECCV paper
- Hacker News thread
What each one is
Hunyuan OCR
Small end-to-end OCR model from Tencent covering parsing, extraction and photo translation.
Hunyuan 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": "tencent/hunyuan-ocr","fallbacks": ["naver/donut"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Hunyuan OCR alternativesDonut alternativesBest OCR models in 2026