Abstract: The development of large language models in the field of Traditional Chinese Medicine (TCM) using artificial intelligence has significantly contributed to both the innovation and preservation of TCM. This paper outlines the current research status and process of large language models in TCM, focusing on three key tasks: collecting TCM data, fine-tuning models with specific instructions, and using different methods to evaluate model performance. It also highlights cutting-edge techniques, such as prompt engineering, retrieval-augmented generation, and reinforcement learning from human feedback, which have enhanced the models' adaptability in various TCM applications. The challenges faced by TCM language models, such as data privacy, ethical biases, model interpretability, technical difficulties, and evaluation standards, are also analyzed, indicating areas for further improvement. Looking ahead, the combination of these models with advanced AI techniques like deep learning, and the integration of multimodal information such as TCM diagnostic data and herbal images, can open new possibilities for large language models in TCM. This will enhance their application in areas like syndrome differentiation diagnosis, prescription recommendations, TCM knowledge graph construction, and TCM education.