27B text-only medical variant
According to the model card, this is the 27B text-only MedGemma variant, trained exclusively on medical text and available only as an instruction-tuned model.
Open Source Model Profile · google
medgemma-27b-text-it
medgemma-27b-text-it is a 27.01B-parameter Gemma 3 text-generation model from Google for medical text comprehension. According to the model card, the 27B variant is text-only and instruction-tuned.
- Publisher
- Task
- text-generation
- Model type
- gemma3_text
- License
- health-ai-developer-foundations
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
medgemma-27b-text-it is published by Google as a Gemma 3 text-generation model for medical text comprehension. The captured configuration identifies Gemma3ForCausalLM with model type gemma3_text, and Safetensors metadata reports 27,009,002,240 parameters. According to the model card, MedGemma variants target healthcare-based AI applications, and the 27B variant is text-only and available only as an instruction-tuned model.
Recorded capabilities
Gemma 3 decoder-only design
According to the model card, the model uses the Gemma 3 decoder-only Transformer architecture with grouped-query attention and supports at least 128K tokens of context.
Documented local and scale paths
The model card provides code snippets for local GPU runs, recommends Model Garden for production scale, and records version 1.0.0 created May 20, 2025.
Captured 27.01B Transformers weights
Safetensors metadata reports 27,009,002,240 parameters and the hub lists Transformers support for text generation.
Use cases in the source record
- Healthcare text-comprehension work, such as medical question or prompt inputs within the documented 128K-token input budget, following the publisher's documentation.
- Local GPU experimentation using the model card's code snippets, with Model Garden recommended by the publisher for production-scale serving.
Limitations and unknowns
- No evaluation scores were extracted from this record.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
- Training, capability, and specification claims are publisher model-card statements and were not independently verified by Ethen.
Source and provenance
Source: google/medgemma-27b-text-it
Captured: Unknown. Processed: 2026-09-07T19:35:22.542366+00:00.
MedGemma model card Model documentation: MedGemma Resources: Model on Google Cloud Model Garden: MedGemma Model on Hugging Face: MedGemma GitHub repository (supporting code, Colab notebooks, discussions, and issues): MedGemma Quick start notebook: GitHub Fine-tuning notebook: GitHub Patient Education Demo built using MedGemma Support: See Contact License: The use of MedGemma is governed by the Health AI Developer Foundations terms of use . Author: Google Model information This section describes the MedGemma model and how to use it. Description MedGemma is a collection of Gemma 3 variants that are trained for performance on medical t…
F001F002F003F004F005F006F007F008F010F011F013F014F015F016F017