LOADING...
Google unveils MedGemma 1.5, MedASR AI models: What's so special?
The move is part of the tech giant's ongoing efforts in the healthcare sector

Google unveils MedGemma 1.5, MedASR AI models: What's so special?

Jan 14, 2026
07:49 pm

What's the story

Google has unveiled two new artificial intelligence (AI) models, MedGemma 1.5 and MedASR, to enhance the analysis of medical images and clinical speech data. The move is part of the tech giant's ongoing efforts in the healthcare sector. Unlike OpenAI's ChatGPT for Healthcare, which is a paid offering for enterprises, Google has opted for a more community-oriented approach by making these models open-source.

Model details

MedGemma 1.5: A leap in medical image analysis

MedGemma 1.5, the latest iteration of Google's open medical vision-language model, is capable of analyzing medical images and text simultaneously. The advanced version improves upon its predecessors with enhanced multimodal reasoning, improved handling of complex medical imagery, and greater flexibility for fine-tuning on specialized datasets. It can process various types of medical images such as radiology scans and other clinically relevant visuals.

Usage caution

Potential applications and limitations

Google has made it clear that MedGemma 1.5 is not meant to provide diagnoses or treatment recommendations. The model is intended as a supporting tool for research and development purposes, with potential use cases including image-based question answering, report generation, and structured data extraction.

Advertisement

Speech model

MedASR: A breakthrough in clinical speech transcription

Along with MedGemma 1.5, Google also unveiled MedASR, an automatic speech recognition model for healthcare. The advanced system is designed to transcribe spoken clinical conversations into text while accurately handling medical terminology, accents, and real-world clinical audio conditions. Google claims that MedASR aims to minimize errors often seen in general-purpose speech recognition systems when applied to healthcare use cases.

Advertisement

Model accessibility

MedASR's adaptability and access

MedASR can be used for transcribing doctor-patient interactions, clinical notes, as well as dictated reports. It is designed to be adaptable across different healthcare environments, and can be fine-tuned for specific clinical workflows/documentation standards. Google has made all variants of MedGemma and the MedASR model available through its Hugging Face listing or Vertex AI platform. The tech giant's MedGemma GitHub repository is also accessible for developers interested in tutorials.

Advertisement