# Google-Health/medgemma

Repository: https://github.com/Google-Health/medgemma
Canonical: https://ross.abutalabs.com/products/medgemma
Language: Jupyter Notebook
License: Apache-2.0
License Family: permissive
Last push: 2026-06-19T17:29:02+00:00

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 88, release rhythm 35, longevity 35
- inputs: {"age_days": 490, "days_push": 75, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1574, forks 269 (observed 2026-08-28T04:05:05.822009+00:00)

## What it is
MedGemma is a collection of Gemma 3 model variants trained on medical text and imagery (a 4B multimodal version and a 27B text-only version) released by Google Health. The repository hosts Jupyter notebooks, documentation links, and evaluation guidance for developers building healthcare AI applications with these models.

## Use cases
- build ai apps that understand medical images like chest x-rays
- analyze dermatology or histopathology images with a multimodal llm
- run a medically-tuned language model for clinical text comprehension
- fine-tune a medical foundation model for a healthcare task
- evaluate llm performance on clinically relevant benchmarks
- get started with medgemma notebooks on hugging face or vertex ai

## When to choose
- you need open model weights pretrained on medical text and imaging data
- you want a starting checkpoint to fine-tune for healthcare applications
- you want guided notebooks for medical multimodal inference

## When to avoid
- you need a production-ready clinical decision system without additional fine-tuning and validation
- you need a general-purpose llm with no medical focus
- you cannot accept the Health AI Developer Foundations license terms for the model weights

## Facets
- artifact type: learning-resource
- maturity: active
- function: machine-learning, deep-learning, llm-inference, image-processing, nlp
- domain: healthcare, artificial-intelligence, large-language-models, machine-learning, computer-vision
- platform: python, cloud
- tags: medical-ai, gemma, multimodal, model-weights, jupyter-notebooks, healthcare, fine-tuning, hugging-face, gpu

## Member repositories
- Google-Health/medgemma (main) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:05.822009+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:57:37.889064+00:00, confidence not recorded.
  - readme: https://github.com/Google-Health/medgemma (fetched 2026-08-28T04:05:05.822009+00:00, sha b41c24f6594b)
- Data as of 2026-08-30T08:39:29.467469+00:00.