PaddleOCR (PP-OCR) vs Hunyuan OCR
Pick PaddleOCR (PP-OCR) for fast line-level text reading on Chinese and English printed text, receipts, signs and screenshots, especially on CPU, edge devices or in the browser. Pick Hunyuan OCR when you need Markdown, HTML, LaTeX or JSON output, information extraction, image translation, or low-resource and ancient scripts from one ~1B model. Check the Hunyuan licence first: it does not apply in the EU, UK or South Korea, while PaddleOCR is Apache-2.0.
PaddleOCR (PP-OCR)paddle/paddleocr | Hunyuan OCRtencent/hunyuan-ocr | |
|---|---|---|
| Type | Open source | Open source |
| Price | 1 credit / page | 2 credits / page |
| Per 1,000 pages | $1.00 | $2.00 |
| Typical speed | 0.7 s / page | 1.5 s / page |
| Best for | Fast printed text, Chinese and English | Chinese documents and photos |
| Inputs | PDF, images | PDF, images |
| Languages | 80+ languages | 100+ languages |
| Max pages per request | 500 | 500 |
| Tables | No | ✓ Yes |
| Handwriting | No | No |
| Math | No | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | ✓ Yes | No |
| Forms | No | No |
| Multilingual | ✓ Yes | ✓ Yes |
| 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 (PP-OCR) if
- It is Apache-2.0, so there are no regional exclusions or user-count caps.
- PP-OCRv6 medium is 34.5M parameters and runs on CPU, NVIDIA GPU, Kunlunxin XPU and other accelerators, with a 5.2x CPU speedup using OpenVINO.
- One PP-OCRv6 model covers 50 languages, including Chinese, English, Japanese and 46 Latin-script languages, and the PaddleOCR.js SDK runs it in the browser.
Choose Hunyuan OCR if
- It outputs Markdown, HTML, LaTeX and JSON, so it can parse documents without a separate table or formula pipeline.
- One ~1B model handles document parsing, text spotting, information extraction and image translation, with 12 task types exposed via the vLLM client.
- It supports images up to 4K and a 128K context window in v1.5, and covers low-resource and ancient-script OCR.
Key differences
| PaddleOCR (PP-OCR) | Hunyuan OCR | |
|---|---|---|
| Licence | Apache-2.0 | Tencent Hunyuan Community License; excludes the EU, UK and South Korea; separate licence needed above 100M monthly active users |
| Model size | PP-OCRv6 medium: 34.5M parameters | About 1B parameters |
| How you run it | paddleocr Python package, C++ or the PaddleOCR.js browser SDK | Self-hosted via vLLM, Transformers or llama.cpp; CUDA 13 recommended for the unified vLLM environment |
| Output formats | Plain text, bounding boxes, JSON; tables and formulas need the separate PP-Structure or VL pipelines | Markdown, HTML, LaTeX, JSON, plain text, bounding boxes |
| Hardware | CPU, NVIDIA GPU, Kunlunxin XPU and other accelerators; 0.13s on A100 reported for PP-OCRv6 | NVIDIA GPU with CUDA 13 for vLLM; llama.cpp GGUF path for CPU, consumer GPU or laptop |
| Benchmark | Not published | ParseBench-2 overall 67.11 and ParseBench mean 63.86 (v1.5, independent) |
| Known failure | PP-OCRv6 weights contain denormals that make CPU inference 4 to 11x slower in RapidOCR; open issues on PaddleOCR 3.7.0 and Windows training | Archived 1.0 note said Transformers and vLLM outputs differed in accuracy; the unified environment needs CUDA 13 and vLLM 0.25.1+ |
Sources
- PaddleOCR GitHub README
- RapidOCR issue #743 denormal weights
- HunyuanOCR GitHub
- HunyuanOCR model card
- Hunyuan Community License
What each one is
PaddleOCR (PP-OCR)
Baidu's classic open-source text detection and recognition pipeline. Light enough to run on CPU.
PaddleOCR (PP-OCR) detailsHunyuan OCR
Small end-to-end OCR model from Tencent covering parsing, extraction and photo translation.
Hunyuan OCR 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","fallbacks": ["tencent/hunyuan-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
PaddleOCR (PP-OCR) alternativesHunyuan OCR alternativesBest OCR models in 2026