numind/NuExtract-tiny

numind/NuExtract-tiny is a 0.6 billion parameter language model developed by NuMind, based on the Qwen1.5-0.5B architecture. It is specifically fine-tuned for information extraction tasks from text, leveraging a private, high-quality synthetic dataset. This model excels at purely extractive information retrieval, ensuring all output text is present in the original input, and supports structured extraction via JSON templates.