There is overwhelming evidence from academic research and industry benchmarks that domain-specific and task-specific large language models outperform general-purpose LLMs across multiple dimensions: Accuracy, veracity, human preference, and cost.This session presents the results of a double-blind study, in which medical doctors compared John Snow Labs’ healthcare-specific LLMs with OpenAI’s GPT-4o across four popular medical language understanding tasks: Medical text summarization, across a variety of patient notes and report types Open-ended medical question answering, testing out-of-the-box general medical knowledge losed-ended question answering – extracting specific information from a given patient note, such as a patient’s primary diagnosis or disease stage Closed-ended biomedical research – understanding a given research paper abstract