NuExtract for Structured Extraction

In-house text-to-.JSON LLM to identify all types of information and structure it into a schema output. Great for document parsing, RAG and agents.

NuExtract Comes With 0.5B, 3.8B and 7B Parameters

These models are purely extractive.
NuExtract
The base version of the collection.
3.8B parameters.
NuExtract is a version of phi-3-mini, fine-tuned on a private high-quality synthetic dataset.
NuExtract_large
7B parameters.
Also a version of phi-3-mini, also fine-tuned on a private high-quality synthetic dataset.
NuExtract_tiny
The zero-shot version of the collection.
0.5B parameters.
NuExtract-tiny is a version of Qwen1.5-0.5, also fine-tuned on a private high-quality synthetic dataset.
While this model provides good zero-shot performance, it is intended to be fine-tuned on a specific task (>=30 examples).

Your Structured Extraction Use Case is Unique?

Get NuExtract customized by our model experts.