Muhammadidrees/Medgamma27B

TEXT GENERATIONConcurrency Cost:2Model Size:27BQuant:FP8Ctx Length:32kPublished:Oct 2, 2025License:otherArchitecture:Transformer Cold

MedGemma 27B is a 27 billion parameter, text-only instruction-tuned variant of the Gemma 3 model developed by Google, specifically trained for performance on medical text comprehension. It utilizes a decoder-only transformer architecture with grouped-query attention and supports a context length of at least 128K tokens. Optimized for inference-time computation, MedGemma 27B excels in medical knowledge and reasoning tasks, outperforming base Gemma models on health benchmarks like MedQA and MedMCQA.

Loading preview...

MedGemma 27B: Specialized Medical Language Model

MedGemma 27B, developed by Google, is a 27 billion parameter, text-only instruction-tuned model built upon the Gemma 3 architecture. It is specifically trained on a diverse corpus of medical text to enhance performance in healthcare-based AI applications. This model is optimized for inference-time computation and features a decoder-only transformer architecture with grouped-query attention, supporting a substantial context length of at least 128K tokens.

Key Capabilities & Performance

  • Medical Text Comprehension: Trained exclusively on medical text, MedGemma 27B demonstrates strong baseline performance in understanding and processing medical information.
  • Enhanced Medical Benchmarks: It significantly outperforms its base Gemma counterparts across various text-only health benchmarks, including MedQA (89.8%), MedMCQA (74.2%), PubMedQA (76.8%), and MMLU Med (87.0%).
  • Instruction-Tuned: Available as an instruction-tuned model, making it ready for direct application in medical question-answering and assistant roles.
  • Fine-tuning Potential: Developers can fine-tune MedGemma variants with proprietary data for improved performance on specific tasks.

Intended Use Cases

  • Healthcare AI Development: Designed as a foundational model for building downstream healthcare AI applications involving medical text.
  • Medical Question Answering: Capable of providing answers to textual medical questions.
  • Research & Development: Ideal for developers in life sciences and healthcare seeking to adapt and validate AI solutions for clinical or research contexts.

MedGemma 27B is intended to be a starting point for developers, requiring appropriate validation and adaptation for specific use cases, and its outputs should be independently verified.