This research evaluates the effectiveness of medium-sized transformer models for processing medical records compared to large language models (LLMs).

• Medium-sized models demonstrate competitive performance for specific medical tasks
• These models require fewer computational resources while maintaining accuracy
• Models like CamemBERT-bio show promise in handling complex medical terminology
• Research addresses challenges of limited annotated medical datasets

For healthcare organizations with limited AI infrastructure, this study provides evidence that smaller, specialized models can effectively process medical records, making advanced NLP more accessible and deployable in clinical settings.

Are Medium-Sized Transformers Models still Relevant for Medical Records Processing?