OCRRouter

Qianfan OCR vs Hunyuan OCR

Pick Qianfan OCR for self-hosted, multilingual document parsing under a standard Apache-2.0 licence, especially tables, formulas and key information extraction. Pick Hunyuan OCR when you need a roughly 1B model that runs on a laptop or consumer GPU, or when you work with photo text, image translation or historical scripts, and you are outside the EU, UK and South Korea.

Qianfan OCRbaidu/qianfan-ocrHunyuan OCRtencent/hunyuan-ocr
TypeOCR APIOpen source
Price2 credits / page2 credits / page
Per 1,000 pages$2.00$2.00
Typical speed2.0 s / page1.5 s / page
Best forChinese documentsChinese documents and photos
InputsPDF, imagesPDF, images
LanguagesChinese, English and more100+ languages
Max pages per request200500
Tables✓ Yes✓ Yes
Handwriting✓ YesNo
Math✓ Yes✓ Yes
Bounding boxesNo✓ Yes
ConfidenceNoNo
FormsNoNo
Multilingual✓ Yes✓ Yes
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 Qianfan OCR if

  • Apache-2.0 allows commercial use with no regional exclusions and no user-count cap.
  • It scores 93.12 on OmniDocBench v1.5, the top end-to-end score in Baidu's table, and it targets tables and formulas.
  • It covers 192 languages plus handwriting and scene text, and it is also offered hosted on the Baidu Qianfan platform.

Choose Hunyuan OCR if

  • At about 1B parameters it runs on a laptop, CPU or consumer GPU through the llama.cpp GGUF path.
  • One model handles 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 it handles low-resource and ancient-script OCR.

Key differences

Qianfan OCRHunyuan OCR
LicenceApache-2.0, commercial use allowedTencent Hunyuan Community License; excludes the EU, UK and South Korea; separate licence needed above 100M monthly active users
Model size4B (Qwen3-4B language model)About 1B
How you run itSelf-host the open Hugging Face weights, or use the Baidu Qianfan platformSelf-host only, via vLLM, Transformers or llama.cpp
HardwareBenchmarked on one NVIDIA A100 with vLLM; no minimum VRAM stated; not suited to CPU-only deploymentNVIDIA GPU with CUDA 13 recommended for the unified vLLM environment; llama.cpp GGUF path for CPU, consumer GPU or laptop
Output formatsMarkdown, JSON, HTML, LaTeX, bounding boxesMarkdown, HTML, LaTeX, JSON, plain text, bounding boxes
BenchmarksOmniDocBench v1.5 overall 93.12; olmOCR-Bench 79.8; OCRBench 880 (reported by the authors)ParseBench-2 overall 67.11 and ParseBench mean 63.86 on v1.5 (independent); not published on OmniDocBench
Known failure or limit1.024 pages per second on an A100 even with W8A8; thinking mode can emit up to 16,384 tokens, adding latencyUnified environment needs CUDA 13 and vLLM 0.25.1+; archived 1.0 note said Transformers and vLLM outputs differed in accuracy

Sources

What each one is

Qianfan OCR

Baidu's OCR-tuned vision model. Strong on Chinese documents and tables.

Qianfan OCR details

Hunyuan OCR

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

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

Questions

More comparisons

Qianfan OCR alternativesHunyuan OCR alternativesBest OCR models in 2026