Google Releases MedGemma 1.5 with Enhanced Medical Imaging and MedASR Speech Model

January 13, 2026

Details

  • Google AI Developers announced MedGemma 1.5, an updated open model from Google Research, small enough to run offline, with improved performance on medical imaging, text, medical records, and 2D images.
  • MedGemma 1.5 4B supports high-dimensional CT and MRI imaging, whole-slide histopathology, longitudinal chest X-ray interpretation, and extraction from medical lab reports.
  • Released alongside MedASR, an open model for medical automated speech recognition to transcribe healthcare vocabulary and generate prompts for MedGemma.
  • Built on Gemma 3, it shows benchmark gains like 3% accuracy on CT findings (61% vs 58%), 14% on MRI (65% vs 51%), and 35% better anatomical localization in chest X-rays.
  • Access available on Vertex AI, Hugging Face; includes DICOM support, tutorial notebooks for CT, histopathology, and agentic orchestration with tools like Gemini.
  • Improves text tasks: 5% on MedQA (69% vs 64%), 18% on lab report extraction, used by developers like Qmed Asia for clinical guidelines.

Impact

Google's MedGemma 1.5 advances open-source medical AI by introducing support for high-dimensional imaging like 3D CT and MRI volumes alongside text, positioning it as one of the first public multimodal models capable of this while handling 2D data and records, potentially pressuring proprietary systems from rivals like Microsoft's Nuance or Nuance-integrated tools in healthcare. This lowers barriers for developers in resource-constrained settings, enabling offline deployment and fine-tuning for privacy-sensitive applications such as local EHR parsing before cloud queries, which aligns with growing regulatory emphasis on data sovereignty under frameworks like HIPAA and GDPR. By integrating with agentic workflows and FHIR tools, it accelerates adoption in clinical decision support, as seen with Qmed Asia's guideline interface, and could shift R&D toward multimodal agents that synthesize imaging time series with speech transcripts via MedASR. Over the next 12-24 months, expect increased funding into specialized health AI startups leveraging these foundations, narrowing gaps in global diagnostic access where 4.7 billion lack imaging services, though clinical validation remains essential to reach production-grade reliability.

Read more