NuExtract3 released: open-weight 4B VLM for Markdown, OCR and structured extraction (self-hostable) [P]

“`html

A new 4B VLM (Very Large Model) named NuExtract3 has been released by Numind. This model is designed for extracting structured data from complex documents like PDFs, screenshots, forms, tables, and receipts.

• It supports a wide range of document types including visually structured inputs such as PDFs and images.
• The model can convert these documents into Markdown format or extract structured JSON data based on predefined templates.
• For optimal performance, the model recommends processing documents page by page for better inference speed when using NuExtract3 for Markdown conversion.

NuExtract3 is available as a self-hostable solution with detailed documentation and multiple quantization options provided. It can be run on systems with minimal VRAM requirements, making it accessible for various use cases.

“`

### Takeaways
– **New Open-Weight 4B Model**: NuExtract3 offers an open-weight approach to document extraction, providing a practical tool for handling complex documents.
– **Versatile Use Cases**: The model supports a variety of document types including PDFs and images, making it useful in multiple applications like invoice processing or form filling.
– **Self-Hosting Option**: Users can easily set up the model locally without needing access to cloud infrastructure.

Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.

Please enable JavaScript in your browser to complete this form.

Email Name

Name *

First

Last

Email *

Submit