MedGemma

GoogleOpen SourceVerified

The MedGemma collection contains Google's most capable open models for medical text and image comprehension, built on Gemma 3. Developers can use MedGemma to accelerate building healthcare-based AI applications. MedGemma comes in two variants: a 4B multimodal version and a 27B text-only version. The 4B version utilizes a SigLIP image encoder specifically pre-trained on de-identified medical data including chest X-rays, dermatology images, ophthalmology images, and histopathology slides.

2025-05-20

4B, 27B

Decoder-only Transformer (based on Gemma 3)

Health AI Developer Foundations terms of use

Compare with other models

Specifications

Parameters: 4B, 27B
Architecture: Decoder-only Transformer (based on Gemma 3)
License: Health AI Developer Foundations terms of use
Context Window: 128,000 tokens
Max Output: 8,192 tokens
Training Data Cutoff: 2025-05-20
Type: multimodal
Modalities: textimage

Benchmark Scores

medqa-4op87.7

medmcqa74.2

pubmedqa76.8

mmlu-med87

medxpertqa26.7

afrimed-qa84

mimic-cxr-f188.9

chexpert-f148.1

dermmcqa-accuracy71.8

slakevqa-f162.3

vqa-rad-f149.9

pathmcqa-accuracy69.8

mimic-cxr-radgraph29.5

See all benchmarks

Advanced Specifications

Model Family: Gemma
Finetuned From: Gemma 3
API Access: Available
Chat Interface: Not Available
Multilingual Support: No
Variants: 4B instruction-tuned (medgemma-4b-it)4B pre-trained (medgemma-4b-pt)27B text instruction-tuned (medgemma-27b-text-it)
Hardware Support: CUDATPUCPU

Capabilities & Limitations

Capabilities: medical image classificationmedical image interpretationchest X-ray interpretationdermatology image analysishistopathology analysisophthalmology image analysisfundus image analysisradiology report generationmedical visual question answeringmedical text comprehensionclinical reasoningmedical knowledge retrievalmultimodal medical understandingpatient interviewingmedical triagingclinical decision supportmedical summarizationprompt engineering adaptationfine-tuning capabilityagentic orchestration
Known Limitations: Not intended for direct clinical diagnosis or treatment recommendationsDeveloper model that requires validation on intended use caseBaseline performance may need improvement through adaptationNot clinical grade without additional fine-tuningPrimarily evaluated on single-image tasksNot evaluated for multi-turn applicationsMay be more sensitive to specific prompts than base Gemma 327B model is text-only without image supportEvaluation primarily in English language
Notable Use Cases: medical image classificationmedical image interpretation and reportingradiology image analysisdigital pathology classificationfundus image analysisskin image classificationmedical visual question answeringpatient interviewing systemsmedical triaging applicationsclinical decision support toolsmedical summarizationhealthcare AI application developmentagentic medical systemsFHIR data processingprivate health data parsingmedical education platforms
Function Calling Support: Yes
Tool Use Support: Yes

Resources

Related Models

Gemini 2.5 Flash

Google

Improved across key benchmarks for reasoning, multimodality, code and long context while getting even more efficient. Best for fast performance on complex tasks.

Gemini Diffusion

Google

Our state-of-the-art, experimental text diffusion model that uses diffusion techniques to explore a new kind of language model that gives users greater control, creativity, and speed in text generation.

Gemma 3n

Google

Gemma 3n is a mobile-first AI model built on a groundbreaking architecture optimized for on-device performance. It features Per-Layer Embeddings (PLE) for reduced RAM usage, allowing larger models to run on mobile devices with a memory footprint comparable to much smaller models. Supports multimodal understanding including text, images, audio, and video.