DeepSeek Open-Source LLMs Rival GPT-4o in Clinical Decision-Making

In a pre-publication study in Nature Medicine, researchers benchmarked the performance of two open-source large language models—DeepSeek-V3 and DeepSeek-R1—on clinical decision support tasks, comparing them to proprietary models like GPT-4o and Gemini 2.0 Flash. Using 125 diverse patient cases spanning both common and rare diseases, the study found that DeepSeek models performed on par with, and sometimes exceeded, the accuracy of their commercial counterparts. Crucially, DeepSeek models can be locally deployed within healthcare institutions, offering a privacy-compliant alternative that aligns with strict data protection regulations. This makes them a scalable and secure choice for real-world clinical applications, paving the way for broader adoption of AI in healthcare without compromising patient confidentiality.