The article introduces SkinGPT-4, an interactive dermatology diagnostic system based on multimodal large language models (LLMs). SkinGPT-4 is designed to enhance the accuracy and efficiency of dermatological diagnoses by leveraging pre-trained vision transformers and LLMs, specifically the Llama-2-13b-chat model. The system is trained on a comprehensive dataset of 52,929 skin disease images, clinical concepts, and doctors' notes, using a two-step training strategy. This approach ensures that the model can recognize medical features in images and provide detailed, natural language descriptions of skin conditions. Users can upload their own skin photos, which are then evaluated by SkinGPT-4, which identifies characteristics, performs in-depth analysis, and provides interactive treatment recommendations. The system's local deployment capability and commitment to user privacy make it a valuable tool for both patients and dermatologists, particularly in underserved regions. Quantitative evaluations on 150 real-life cases, reviewed by board-certified dermatologists, demonstrated that SkinGPT-4 consistently provided accurate diagnoses, with high agreement and usefulness in identifying disease types and suggesting treatments. The study highlights the potential of multimodal LLMs in advancing dermatological diagnosis and improving healthcare access and equity.The article introduces SkinGPT-4, an interactive dermatology diagnostic system based on multimodal large language models (LLMs). SkinGPT-4 is designed to enhance the accuracy and efficiency of dermatological diagnoses by leveraging pre-trained vision transformers and LLMs, specifically the Llama-2-13b-chat model. The system is trained on a comprehensive dataset of 52,929 skin disease images, clinical concepts, and doctors' notes, using a two-step training strategy. This approach ensures that the model can recognize medical features in images and provide detailed, natural language descriptions of skin conditions. Users can upload their own skin photos, which are then evaluated by SkinGPT-4, which identifies characteristics, performs in-depth analysis, and provides interactive treatment recommendations. The system's local deployment capability and commitment to user privacy make it a valuable tool for both patients and dermatologists, particularly in underserved regions. Quantitative evaluations on 150 real-life cases, reviewed by board-certified dermatologists, demonstrated that SkinGPT-4 consistently provided accurate diagnoses, with high agreement and usefulness in identifying disease types and suggesting treatments. The study highlights the potential of multimodal LLMs in advancing dermatological diagnosis and improving healthcare access and equity.