medgemma-27b-it
by google
Gemma-based vision-language model tuned for medical imaging and report generation
google/medgemma-27b-itmixpeek://image_extractor@v1/medgemma_27b_it_v1Overview
MedGemma 27B (instruction-tuned) is Google's medical adaptation of Gemma 3, taking image and text input and producing text. Its Hugging Face card tags the modalities it targets directly: x-ray, chest x-ray, pathology, dermatology, fundus, radiology report generation and medical embeddings.
On Mixpeek it fits the same slot as a general VLM in a frames pipeline, with a domain vocabulary that a general model does not have. It is released under a custom licence rather than Apache, so read the terms before deploying it, particularly for anything clinical.
Architecture
Gemma 3 architecture, 27B parameters, instruction-tuned, image-text-to-text. Distributed as safetensors for transformers.
Mixpeek SDK Integration
import { Mixpeek } from "mixpeek";
const mixpeek = new Mixpeek({ apiKey: process.env.MIXPEEK_API_KEY });
await mixpeek.collections.create({
collection_name: "imaging_studies",
feature_extractors: [
{ extractor: "mixpeek://image_extractor@v1/medgemma_27b_it_v1" },
],
});Capabilities
- Radiology report generation from imaging studies
- Visual question answering with medical vocabulary
- Pathology, dermatology and fundus image description
- Medical embeddings for domain retrieval
Use Cases on Mixpeek
Explore on Mixpeek
Compare alternatives in this category
Hand-picked tools & platforms compared
Deep-dive technical guide
See how Mixpeek runs models as extractors
Store & search embeddings at scale
Usage-based pricing for pipelines
Compare models, APIs & infrastructure
Specification
Research Paper
MedGemma 27B IT on Hugging Face
arxiv.orgBuild a pipeline with medgemma-27b-it
Add this model to a processing pipeline alongside other extractors. Combine with retrieval stages for end-to-end search.
Run it on your own data, free