MedGemma Technical Report
cs.AI, cs.CL, cs.CV
Submitted: 2025-07-07
Updated: 2026-04-06
Comments: Fix references
Journal ref: Nat. Med. 32, 3695-3704 (2026)
DOI: 10.1038/s41591-026-04626-w
Code: https://github.com/Google-Health/google-health
License: http://creativecommons.org/licenses/by/4.0/
The gist: Artificial intelligence (AI) has significant potential in healthcare applications, but its training and deployment faces challenges due to healthcare's diverse data, complex tasks…
Terminology
Abstract
Artificial intelligence (AI) has significant potential in healthcare applications, but its training and deployment faces challenges due to healthcare's diverse data, complex tasks, and the need to preserve privacy. Foundation models that perform well on medical tasks and require less task-specific tuning data are critical to accelerate the development of healthcare AI applications. We introduce MedGemma, a collection of medical vision-language foundation models based on Gemma 3 4B and 27B. MedGemma demonstrates advanced medical understanding and reasoning on images and text, significantly exceeding the performance of similar-sized generative models and approaching the performance of task-specific models, while maintaining the general capabilities of the Gemma 3 base models. For out-of-distribution tasks, MedGemma achieves 2.6-10% improvement on medical multimodal question answering, 15.5-18.1% improvement on chest X-ray finding classification, and 10.8% improvement on agentic evaluations compared to the base models. Fine-tuning MedGemma further improves performance in subdomains, reducing errors in electronic health record information retrieval by 50% and reaching comparable performance to existing specialized state-of-the-art methods for pneumothorax classification and histopathology patch classification. We additionally introduce MedSigLIP, a medically-tuned vision encoder derived from SigLIP. MedSigLIP powers the visual understanding capabilities of MedGemma and as an encoder achieves comparable or better performance than specialized medical image encoders. Taken together, the MedGemma collection provides a strong foundation of medical image and text capabilities, with potential to significantly accelerate medical research and development of downstream applications. The MedGemma collection, including tutorials and model weights, can be found at https://goo.gle/medgemma.
Sources
- GPT-4 Technical Report
- Medical Large Language Model Benchmarks Should Prioritize Construct Validity
- MAIRA-2: Grounded Radiology Report Generation
- MedHELM: Holistic Evaluation of Large Language Models for Medical Tasks
- Big Self-Supervised Models are Strong Semi-Supervised Learners
- PaLI: A Jointly-Scaled Multilingual Language-Image Model
- Generating Radiology Reports via Memory-driven Transformer
- UniChest: Conquer-and-Divide Pre-training for Multi-Source Chest X-Ray Classification
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- RadVLM: A Multitask Conversational Vision-Language Model for Radiology
- Gemma 3 Technical Report
- MedAlpaca -- An Open-Source Collection of Medical Conversational AI Models and Training Data
- Measuring Massive Multitask Language Understanding
- MAIRA-1: A specialised large multimodal model for radiology report generation
- RadGraph: Extracting Clinical Entities and Relations from Radiology Reports
- PubMedQA: A Dataset for Biomedical Research Question Answering
- Health AI Developer Foundations
- BioMistral: A Collection of Open-Source Pretrained Large Language Models for Medical Domains
- Improved Baselines with Visual Instruction Tuning
- Improving Factual Completeness and Consistency of Image-to-Text Radiology Report Generation
Related papers
- Multi-Objective Structured Pruning of LLMs for Latency and Model Size Optimization
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Embeddings for Preferences, Not Semantics
- Can We Trust Item Response Theory for AI Evaluation?
- Diagnosing Task Insensitivity in Language Agents