This paper introduces a novel approach to materials classification using large language models (LLMs), prompt engineering, and deep learning (DL). The authors develop a general workflow that leverages LLMs to generate textual data, which is then used to train a bidirectional encoder representations from transformers (BERT) model for classification tasks. The workflow is applied to the classification of metallic glasses (MGs), achieving up to 463% improvement in prediction accuracy compared to conventional classification models. The study highlights the potential of using textual knowledge generated by LLMs to address materials classification problems, especially in scenarios with sparse datasets. The workflow is flexible and can be extended to various material applications, demonstrating its effectiveness in material discovery and design. The authors also provide a detailed methodology, including data processing, LLM and prompt engineering, BERT training, and interpretability analysis, to support their findings.This paper introduces a novel approach to materials classification using large language models (LLMs), prompt engineering, and deep learning (DL). The authors develop a general workflow that leverages LLMs to generate textual data, which is then used to train a bidirectional encoder representations from transformers (BERT) model for classification tasks. The workflow is applied to the classification of metallic glasses (MGs), achieving up to 463% improvement in prediction accuracy compared to conventional classification models. The study highlights the potential of using textual knowledge generated by LLMs to address materials classification problems, especially in scenarios with sparse datasets. The workflow is flexible and can be extended to various material applications, demonstrating its effectiveness in material discovery and design. The authors also provide a detailed methodology, including data processing, LLM and prompt engineering, BERT training, and interpretability analysis, to support their findings.