Google Launches MedGemma 1.5 & MedASR to Broaden Healthcare AI Capabilities

The models are part of Google’s Health AI Developer Foundations programme and are intended to help developers build, test, and scale healthcare applications.

Google has launched MedGemma 1.5 and MedASR, two new healthcare-focused artificial intelligence models aimed at advancing medical image analysis and clinical speech-to-text capabilities, reinforcing the company’s expanding role in the rapidly growing digital health ecosystem.

Google has expanded its healthcare AI portfolio, making both models openly available for research and commercial use through platforms such as Hugging Face and Google Cloud’s Vertex AI.

The models are part of Google’s Health AI Developer Foundations programme and are intended to help developers build, test, and scale healthcare applications while encouraging careful validation before real-world deployment.

According to Google, the industry is embracing artificial intelligence at nearly twice the pace of the broader economy, driven by growing clinical workloads, complex data environments, and the need for more efficient diagnostic and documentation tools.

Last year, Google introduced the original MedGemma model as an open starting point for medical AI development, and the response, the company said, “has been incredible.”

While earlier versions of MedGemma 1.5 focused largely on two-dimensional medical images, the new model supports high-dimensional medical imaging, including three-dimensional CT scans, MRI volumes, and whole-slide pathology images.

Developers can input multiple image slices or patches, allowing the model to work with richer and more complex clinical data.

In addition to imaging, MedGemma 1.5 has shown improved performance on medical text tasks such as electronic health record interpretation and question answering, supported by new training techniques and datasets. Google Cloud deployments of the model also include full DICOM compatibility, aligning it more closely with clinical imaging standards.

Alongside MedGemma 1.5, Google has unveiled MedASR, a speech-to-text model built specifically for healthcare environments.

MedASR has been trained on healthcare-specific language and fine-tuned to handle clinical dictation, medical terminology, accents, and challenging audio conditions common in hospitals and clinics.

The output from MedASR can also be integrated into downstream AI workflows, including multimodal systems that combine speech, text, and medical images. Google has emphasized that MedGemma 1.5 is not intended to provide diagnoses or treatment recommendations.

Instead, both models are positioned as foundational tools for research, development, and workflow support.

Stay tuned for more such updates on Digital Health News