Overview

MedTurk-MedGemma-4b is a 4.3 billion parameter language model developed by ASAIs-TDDI-2025. It is based on the Gemma3 architecture and was finetuned from unsloth/medgemma-4b-it-unsloth-bnb-4bit. The finetuning process utilized Unsloth and Huggingface's TRL library, which allowed for a significantly accelerated training time, reportedly 2x faster.

Key Capabilities

• Efficient Training: Leverages Unsloth for faster finetuning, making it more resource-efficient for developers.
• Gemma3 Architecture: Built upon the robust Gemma3 foundation, providing strong general language understanding and generation capabilities.
• General Purpose: Suitable for a wide range of natural language processing tasks due to its foundational model and finetuning approach.

Good For

• Developers looking for a Gemma3-based model with optimized training efficiency.
• Applications requiring a 4.3 billion parameter model for various language-related tasks.
• Experimentation with models finetuned using Unsloth and TRL for performance benefits.