A free starting point for health-tech developers, but Google says it is not clinical-grade and needs validation before use.
Kenning's take
What it does
MedGemma is a collection of Google open-weight models for medical text and image understanding, built on Gemma 3. Developers use it as a starting point for tasks such as drafting reports on X-rays and scans, answering questions about medical images, summarising records and supporting clinical reasoning. Google says it is not clinical-grade and must be validated and usually fine-tuned before use; the weights are free to download under the Health AI Developer Foundations terms.
- 4B multimodal and 27B text and multimodal versions
- MedGemma 1.5 4B adds CT, MRI and whole-slide pathology images
- Extraction of structured data from lab reports and health records
- Runs on any cloud, local server or edge device
- Managed fine-tuning in Google Cloud Model Garden
Who it suits. Developers and researchers building healthcare AI applications who can validate and adapt a model for their use case.
Key facts
- Category
- AI models
- Made by
- Based in
- United States
- Launched
- 2025
- How to get it
- free download of open weights (Hugging Face, Google Cloud Model Garden) under the Health AI Developer Foundations terms of use
- Works on
- self-hosted, Google Cloud
- Free plan
- Free open weights under HAI-DEF terms; code under Apache 2.0
FAQ
Is MedGemma free?
Yes. The weights are free to download and use under Google's Health AI Developer Foundations terms.
Is MedGemma open source?
Google calls it open-weight, not open-source. The weights are free to use under its terms, and the accompanying code is Apache 2.0.
Can MedGemma be used for diagnosis?
Google says it is not clinical-grade and needs validation and usually fine-tuning before any production use.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.