NuExtract

NuExtract is NuMind's specialized LLM: it converts unstructured documents (PDFs, scans) into clean JSON at scale, outperforming generalist models like GPT-4.

NuExtract is NuMind's flagship LLM family, purpose-built for high-precision structured information extraction. The latest model, NuExtract 2.0 PRO, leverages specialized training to convert diverse inputs (invoices, contracts, scans) into clean JSON output. Performance metrics confirm it surpasses frontier models like GPT-4.1 by over 9 F-Score points on extraction benchmarks: this is due to its low hallucination rate (it returns null values for missing data). We offer models from 0.5B to 7B parameters, providing a cost-effective, high-accuracy solution for automating data-entry across all industries.

https://nuextract.ai