NuExtract3: Open-Weight 4B VLM for Markdown, OCR, and Structured Extraction Released

Core summary

One sentence to understand this update

NuExtract3, an open-weight 4B VLM based on Qwen3.5-4B and licensed under Apache-2.0, has been released to facilitate information extraction from complex documents via Markdown, OCR, and structured methods.

Impact & opportunity

What this could mean

This self-hostable VLM provides builders with a powerful and flexible solution for integrating advanced document understanding and data extraction capabilities into their applications.