Google drops MedGemma 1.5 and MedASR: Big leap for medical AI
Google just rolled out MedGemma 1.5 and MedASR—two open-source AI models designed to help doctors and researchers handle medical data, from X-rays to audio notes.
You can grab them now on Hugging Face or Vertex AI, making cutting-edge tools more accessible than ever.
What makes these models special?
MedGemma 1.5 handles massive amounts of text and images (think CT scans, MRIs, pathology slides) and outperforms its previous version on key medical tasks like answering health record questions or analyzing lab results.
Meanwhile, MedASR is all about audio—it transcribes chest X-ray dictations with a super low error rate, beating the old Whisper model by up to 82%.
Open-source = more innovation
Both models are free to use, tweak, or combine in your own projects—even if you're just curious or want to build something new for healthcare.
Google's also dropped tutorials on GitHub so anyone can jump in and start experimenting.