Skip to main content
The API uses AI models to extract content from your documents. You can control which model processes your request using the model_type parameter.

​
Available Models

Modelmodel_type valueDescription
Nanonets OCR-3nanonets-ocr-3Latest model. Current default.
Nanonets OCR-2nanonets-ocr-2Previous generation model.

​
Choosing a Model

Production recommendation: Always specify an explicit model_type in production integrations. The default model may change over time as newer models are released. Pinning a specific model ensures consistent behavior.
curl -X POST "https://extraction-api.nanonets.com/api/v1/extract/sync" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "file=@document.pdf" \
  -F "output_format=markdown" \
  -F "model_type=nanonets-ocr-3"

​
When to use each model

nanonets-ocr-3 (recommended)

Best for most use cases. Handles rotated/skewed pages, complex layouts, and multi-page documents with higher accuracy. Uses crop augmentation for dense pages.

nanonets-ocr-2

Use when you need consistent behavior with an existing integration, or if you have validated your pipeline against this model and don’t want to re-test.

​
Default Behavior

When model_type is omitted (or set to empty string / "default"), the API uses the current default model. Today this is nanonets-ocr-3, but it may change in a future release. 
# Uses the current default (nanonets-ocr-3)
-F "output_format=markdown"

# Explicitly pinned — won't change when defaults update
-F "output_format=markdown" -F "model_type=nanonets-ocr-3"