BioMedLM

BioMedLM is a biomedical language model based on the GPT architecture, developed collaboratively by the Stanford University and DataBricks teams. It has 270 million parameters and is trained on specialized knowledge in the biomedical field, enabling it to effectively answer complex questions related to medicine and biology. It can be easily fine-tuned on a single A100 GPU and run inference on a laptop. BioMedLM achieves a score of 57.3% on MedMCQA and 69.0% on the MMLU medical genetics exam. Features include a biomedical question answering system, patient consultation response, literature retrieval and summarization, data privacy and on-premise deployment, and complete record of model training data.