OCRRouter

Pixtral vs Unstructured

Pick Unstructured to ingest mixed files (PDF, HTML, Word, PowerPoint, email, images) into a RAG pipeline, especially born-digital PDFs on the fast strategy. Pick Pixtral only if you already run on the Mistral API and need question answering over charts and documents, since Mistral marks it deprecated. Neither is a clean pick for accurate tables: the open-source Unstructured library scores 0.426 on table cell content against 0.866 hosted.

Pixtralmistral/pixtralUnstructuredunstructured/unstructured
TypeVision modelOCR API
Price2 credits / page20 credits / page
Per 1,000 pages$2.00$20.00
Typical speed3.0 s / page3.0 s / page
Best forQuestions about document imagesMixed file types into RAG
InputsimagesPDF, images, DOCX
LanguagesMultilingualMultilingual
Max pages per request5001,000
TablesNo✓ Yes
HandwritingNoNo
MathNoNo
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 Pixtral if

  • It scores 90.7 ANLS on DocVQA and 81.8 on ChartQA, according to Mistral, so it suits question answering over documents and charts.
  • The 128K token context lets you send many images in one request.
  • If you already call pixtral-12b-2409 on the Mistral API, you keep your integration and pay $0.15 per million input and output tokens (launch price).

Choose Unstructured if

  • One library handles PDFs, HTML, Word, PowerPoint, emails and images and returns typed elements as JSON, HTML or text, with bounding boxes.
  • The hosted API gives you the first 10,000 pages free, then $0.015 per page.
  • The Auto, Fast, High-Res and VLM strategies let you trade speed for quality, and the library comes with many connectors.

Key differences

PixtralUnstructured
What it isVision language model, 12B Mistral Nemo decoder plus 400M vision encoderDocument ETL library and hosted platform that partitions files into typed elements
LicenceApache-2.0 open weightsApache-2.0 library; hosted platform is proprietary
Price$0.15 per million tokens on the Mistral API at launch; free to self-host, you pay for computeHosted: 10,000 pages free, then $0.015 per page; open-source library is free
Output formatsPlain text, Markdown, JSONJSON, HTML, plain text, bounding boxes
BenchmarkDocVQA 90.7 ANLS, ChartQA 81.8, MMMU 52.5 (reported by Mistral)Table cell content 0.426 open-source vs 0.866 hosted (vendor's own benchmark)
Handwriting and non-Latin textReported accuracy drop on handwriting, small fonts and non-Latin scripts; Q4 quantization degrades small textNot published for handwriting; table OCR ignores language settings
Status and known failuresMarked deprecated and no longer maintained by Mistral; a launch-thread user said OCR is not always close to correcthi_res tables can cut off at the edges and drop first letters; extra Tesseract pass on tables adds time; library sends usage telemetry by default

Sources

What each one is

Pixtral

Mistral's open-weight vision models, Pixtral 12B and Pixtral Large. Mistral's API no longer serves them, so they run self-hosted.

Pixtral details

Unstructured

Document ETL for RAG. Partitions many file types into typed elements; open-source library too.

Unstructured 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": "mistral/pixtral","fallbacks": ["unstructured/unstructured"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

Pixtral alternativesUnstructured alternativesBest OCR models in 2026