Google publishes MedGemma, open medical vision-language models

Google has published MedGemma, a family of open vision-language models for medicine, in Nature Medicine. The release spans a 4-billion-parameter multimodal model, a 27-billion-parameter multimodal model, a 27-billion-parameter text-only sibling, and MedSigLIP, a medical image encoder derived from SigLIP. All are built on Google's open Gemma 3 base.

The paper reports that MedGemma "demonstrates advanced medical understanding and reasoning across images and text and multiple medical imaging domains, exceeding the performance of similarly sized generative models while maintaining the general capabilities of the Gemma base models." On out-of-distribution tasks, the authors report "improvements of 2.6–10% in medical image question answering, 15.5–18.1% in chest X-ray finding classification and 10.8% in agentic evaluations compared with the base models."

The framing in the Nature Medicine paper treats access, not raw capability, as the bottleneck in healthcare AI: "Artificial intelligence has high potential for impact in healthcare applications, but its training and deployment are challenging due to diverse data, a complex spectrum of possible tasks and important privacy needs." Open weights are the authors' answer: "High-performing foundation models that enable data-efficient fine-tuning for diverse downstream tasks can meaningfully accelerate development in this domain."

Two of the researchers we track at AI Weekly shared the paper's source on release.