AI Briefing
KO

NuExtract3 Gets Official Ollama Support

·2026.07.22 21:58

Key point

NuMind's open-weight 4B VLM NuExtract3 has been officially released on Ollama, enabling local structured document extraction.

Details

NuExtract3 is an Apache-2.0 open-weight 4B VLM based on Qwen3.5-4B, specialized for document understanding rather than general conversation. It was developed by the NuMind team, and official Ollama support has now been added.

Ollama deployment variants:

  • Q4_K_M: 3.4 GB (recommended for general local use)
  • Q6_K: 4.1 GB (improved precision)
  • BF16: 9.3 GB (original precision)

Key features:

  • Document image or text → JSON template-based structured extraction
  • Document image → clean Markdown conversion
  • Markdown output including HTML tables and LaTeX equations
  • Support for receipts, invoices, forms, contracts, scanned documents, tables, and complex layouts
  • Support for multilingual documents and multiple images/multi-page documents
  • Thinking/Non-thinking reasoning modes

In the Ollama Python client, you can extract structured data from images by passing a JSON schema with role: template.

ollamaqwenvlmlocal-llmopen-weightsdocument-extractionnuextract3structured-json
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

Hacker NewsApple Releases LensVLM-9B for Processing Compressed Text ImagesTLDR AIInvolution and Overseas Expansion of Chinese AI: Structural Limitations and US Response StrategiesRedditVS Code Local LLM Agent Extension UpdateRedditPiper Releases Local Coding Worker for Cloud AgentsTLDR AIQwen Releases 'Qwen-Drive-1.0', a Vision-Language Foundation Model for Autonomous DrivingHacker NewsOllaya Launches Local Runtime for Open-Source Decision Models with TypeSafe SDK Compatibility