This survey explores the application of large language models (LLMs) in medicine, highlighting their potential to address challenges in medical information retrieval, processing, and decision-making. LLMs, based on deep learning, are capable of processing and generating natural language text, and have shown significant improvements in performance across various natural language processing (NLP) tasks. The development of LLMs is driven by the availability of large-scale pre-training datasets and advancements in computational resources. LLMs can effectively understand and reason with medical texts, comprehend medical terminology, and provide accurate retrieval and processing of medical information. They can integrate diverse medical data sources, extract knowledge and insights from vast datasets, and provide personalized recommendations and advice tailored to user needs and preferences. The training of medical LLMs typically involves seven steps: data collection, data preprocessing, model selection and architecture design, model training, hyperparameter tuning, validation and evaluation, and model deployment and application. The process of training a medical LLM involves gathering a large-scale corpus of text data in the medical domain, preprocessing the data, selecting and designing an appropriate model architecture, training the model, and evaluating its performance. The model's performance is evaluated using metrics such as perplexity, text quality, and accuracy. Once the model training and evaluation are complete, they can be deployed in practical applications to fulfill the information needs of medical professionals and patients. LLMs can provide accurate, timely, and reliable medical information, thereby improving the quality and efficiency of medical decision-making. The survey also discusses the characteristics of medicine and the requirements of medical LLMs, including compassionate care, interpretability, practice-oriented, team collaboration, ethical challenges, uncertainty and complexity, and diverse fields. The applications of LLM in medicine include dentistry, radiology, nuclear, clinical, and drug design. The survey also presents various LLM products for medicine, including BenTsao (Huatuo), Med-PaLM, PanGu drug model, Deep Synergistic Interaction Network (DSI-Net), Medical localize and segment anything model (MedLSAM), PubMed GPT, ChatDoctor, and Multimodal LLMs such as Med-MLLM and PeFoMed. These models have shown promising results in various medical applications, including supplementary treatment and diagnosis, drug design, medical image segmentation, doctor-patient communication, and multimodal LLMs. The survey concludes that exploring the applications, limitations, and potential advancements of LLMs in the medical domain is crucial, and this survey aims to provide a comprehensive overview of the utilization of LLMs in medicine, including their benefits, challenges, and emerging trends.This survey explores the application of large language models (LLMs) in medicine, highlighting their potential to address challenges in medical information retrieval, processing, and decision-making. LLMs, based on deep learning, are capable of processing and generating natural language text, and have shown significant improvements in performance across various natural language processing (NLP) tasks. The development of LLMs is driven by the availability of large-scale pre-training datasets and advancements in computational resources. LLMs can effectively understand and reason with medical texts, comprehend medical terminology, and provide accurate retrieval and processing of medical information. They can integrate diverse medical data sources, extract knowledge and insights from vast datasets, and provide personalized recommendations and advice tailored to user needs and preferences. The training of medical LLMs typically involves seven steps: data collection, data preprocessing, model selection and architecture design, model training, hyperparameter tuning, validation and evaluation, and model deployment and application. The process of training a medical LLM involves gathering a large-scale corpus of text data in the medical domain, preprocessing the data, selecting and designing an appropriate model architecture, training the model, and evaluating its performance. The model's performance is evaluated using metrics such as perplexity, text quality, and accuracy. Once the model training and evaluation are complete, they can be deployed in practical applications to fulfill the information needs of medical professionals and patients. LLMs can provide accurate, timely, and reliable medical information, thereby improving the quality and efficiency of medical decision-making. The survey also discusses the characteristics of medicine and the requirements of medical LLMs, including compassionate care, interpretability, practice-oriented, team collaboration, ethical challenges, uncertainty and complexity, and diverse fields. The applications of LLM in medicine include dentistry, radiology, nuclear, clinical, and drug design. The survey also presents various LLM products for medicine, including BenTsao (Huatuo), Med-PaLM, PanGu drug model, Deep Synergistic Interaction Network (DSI-Net), Medical localize and segment anything model (MedLSAM), PubMed GPT, ChatDoctor, and Multimodal LLMs such as Med-MLLM and PeFoMed. These models have shown promising results in various medical applications, including supplementary treatment and diagnosis, drug design, medical image segmentation, doctor-patient communication, and multimodal LLMs. The survey concludes that exploring the applications, limitations, and potential advancements of LLMs in the medical domain is crucial, and this survey aims to provide a comprehensive overview of the utilization of LLMs in medicine, including their benefits, challenges, and emerging trends.