Google has released updates to its MedGemma model, now at version 1.5, enhancing its capabilities in medical image interpretation. This new version supports various imaging modalities, including CT, MRI, and histopathology, allowing developers to input multiple slices or patches for tasks like anatomical localization and disease classification. Performance benchmarks show significant improvements: a 3% increase in accuracy for CT findings and a 14% increase for MRI findings compared to the previous version. The updated model is designed to be compute-efficient, making it suitable for offline use, while a more complex 27B parameter model remains available for heavier applications. In tandem with MedGemma, Google introduced MedASR, an automated speech recognition model fine-tuned for medical dictation. MedASR boasts a word error rate of 5.2% on chest X-ray dictations, significantly outperforming the generalist Whisper model, which has a 12.5% error rate. This model can transcribe medical conversations and generate prompts for MedGemma, facilitating a smoother interaction between healthcare professionals and AI. Both models are available for free under the Health AI Developer Foundations program, encouraging developers to experiment and build new applications. To stimulate innovation, Google launched the MedGemma Impact Challenge on Kaggle, offering $100,000 in prizes for developers who create impactful healthcare solutions using these models. The initiative aims to explore the potential of AI in transforming healthcare, inviting contributions that leverage the advancements in MedGemma and MedASR.