Healthcare is increasingly embracing AI to enhance workflow management, patient communication, and diagnostic and treatment support. It’s critical that these AI-based systems usually are not only high-performing, but in addition efficient and privacy-preserving. It’s with these considerations in mind that we built and recently released Health AI Developer Foundations (HAI-DEF). HAI-DEF is a set of lightweight open models designed to supply developers robust starting points for their very own health research and application development. Because HAI-DEF models are open, developers retain full control over privacy, infrastructure and modifications to the models. In May of this yr, we expanded the HAI-DEF collection with MedGemma, a set of generative models based on Gemma 3 which can be designed to speed up healthcare and lifesciences AI development.
Today, we’re proud to announce two recent models on this collection. The primary is MedGemma 27B Multimodal, which complements the previously-released 4B Multimodal and 27B text-only models by adding support for complex multimodal and longitudinal electronic health record interpretation. The second recent model is MedSigLIP, a light-weight image and text encoder for classification, search, and related tasks. MedSigLIP is predicated on the identical image encoder that powers the 4B and 27B MedGemma models.
MedGemma and MedSigLIP are strong starting points for medical research and product development. MedGemma is helpful for medical text or imaging tasks that require generating free text, like report generation or visual query answering. MedSigLIP is really helpful for imaging tasks that involve structured outputs like classification or retrieval. The entire above models will be run on a single GPU, and MedGemma 4B and MedSigLIP may even be adapted to run on mobile hardware.
Full details of MedGemma and MedSigLIP development and evaluation will be present in the MedGemma technical report.
