Abstract: The advent of medical large language models (LLMs) has created new opportunities for intelligent cancer diagnosis. Compared to traditional clinical decision support systems that rely on empirical knowledge and rule⁃based methodologies, LLMs possess the capability to capture intricate semantic representations and latent relationships within multimodal clinical data. The large⁃model technology framework is driving the intelligent transformation in tumor diagnosis, enhancing the accuracy and logical reasoning of clinical decisions, improving the organization of medical knowledge, and strengthening physicians’ analytical capabilities in complex cases. Despite persistent challenges in interpretability, reliability, and data privacy persist, these models are establishing a robust foundation for the clinical realization of intelligent medicine. This article summarizes critical  advances in the application of medical LLMs for cancer diagnosis, including progress in model architecture optimization, domain adaptation, and multimodal data integration. It explores the potential in key scenarios such as clinical information extraction, joint imaging⁃pathology analysis, and digital diagnostic support, providing systematic reference for cancer research and clinical practice.

Key words: Cancer diagnosis, Clinical decision support systems, Medical large language models, Multimodal data integration,  ,
Digital diagnostic support