Abstract: With the rapid development of artificial intelligence technologies, large language models (LLMs), leveraging their advantages in semantic understanding, logical reasoning, and cross-task generalization, are gradually becoming an important technological direction for promoting the intelligent upgrading of the steel industry.Focusing on the construction and application of LLMs in the steel industry, this paper starts from the development background and technical characteristics of industrial large models and systematically reviews their research progress and representative applications across the entire steel production process.Considering the characteristics of the steel industry, including dense multi-source heterogeneous data, complex operating conditions, and long decision-making chains, this study organizes relevant methodological frameworks and intrinsic logical relationships from three key perspectives: data processing and dataset construction, fault diagnosis and analysis, and process scheduling optimization. It further elucidates the technical pathway through which LLMs evolve from auxiliary analytical tools toward comprehensive intelligent decision-support roles across the full production workflow.On this basis, key challenges such as multimodal data alignment, model reliability, and interpretability are summarized and analyzed, and potential application directions of LLMs in steel production are discussed, providing a reference for systematic research and engineering applications of LLMs in the steel industry.