Hunyuan OCR vs Pixtral
Pick Hunyuan OCR if you can self-host and need scene text spotting, image translation, or low-resource and historical scripts, and you are not in the EU, UK or South Korea. Pick Pixtral only if you already run on the Mistral API or need question answering over charts and documents. Pixtral is deprecated and unmaintained, so it is a poor base for a new OCR project.
Hunyuan OCRtencent/hunyuan-ocr | Pixtralmistral/pixtral | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 1.5 s / page | 3.0 s / page |
| Best for | Chinese documents and photos | Questions about document images |
| Inputs | PDF, images | images |
| Languages | 100+ languages | Multilingual |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | No |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | No | 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 Hunyuan OCR if
- It is a ~1B model that handles parsing, text spotting, information extraction and image translation, and it runs on a laptop through llama.cpp GGUF.
- It accepts images up to 4K with a 128K context window in v1.5, and exposes 12 task types including low-resource and ancient-script OCR.
- It outputs Markdown, HTML, LaTeX, JSON, plain text and bounding boxes, so you get layout data as well as text.
Choose Pixtral if
- It has Apache-2.0 open weights, with no regional exclusions and no user-count cap.
- You can call it through the Mistral API at a launch price of $0.15 per million input and output tokens, or self-host it.
- It scores 90.7 ANLS on DocVQA and 81.8 on ChartQA (Mistral's own numbers), and its 128K context fits many images per request.
Key differences
| Hunyuan OCR | Pixtral | |
|---|---|---|
| Licence | Tencent Hunyuan Community License; excludes the EU, UK and South Korea; separate license above 100M monthly active users | Apache-2.0 |
| How you run it | Self-host only, via vLLM, Transformers or llama.cpp | Mistral API (pixtral-12b-2409) or self-host open weights |
| Price | Free to self-host; you pay for compute | $0.15 per million input and output tokens at API launch; free to self-host |
| Size | ~1B parameters | 12.4B (12B decoder + 400M vision encoder) |
| Output formats | Markdown, HTML, LaTeX, JSON, plain text, bounding boxes | Plain text, Markdown, JSON |
| Benchmarks | ParseBench-2 overall 67.11; ParseBench mean 63.86 (v1.5, independent) | DocVQA 90.7 ANLS, ChartQA 81.8, MMMU 52.5 (from Mistral) |
| Known failures | Archived 1.0 note: Transformers and vLLM outputs differed in accuracy; unified setup needs CUDA 13 and vLLM 0.25.1+ | Deprecated by Mistral; weaker on handwriting, small fonts and non-Latin scripts; Q4 quantization degrades small text OCR |
Sources
- HunyuanOCR model card
- Hunyuan Community License
- HunyuanOCR GitHub
- Mistral announcement
- Mistral Sept 2024 pricing
- Third-party deployment guide
What each one is
Hunyuan OCR
Small end-to-end OCR model from Tencent covering parsing, extraction and photo translation.
Hunyuan OCR detailsPixtral
Mistral's open-weight vision models, Pixtral 12B and Pixtral Large. Mistral's API no longer serves them, so they run self-hosted.
Pixtral 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": ["mistral/pixtral"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Hunyuan OCR alternativesPixtral alternativesBest OCR models in 2026