From gigabytes to the bedside: Small language models as practical clinical AI in anaesthesiology

The rapid rise of large language models (LLMs) has dominated contemporary discourse around artificial intelligence (AI) in medicine.By taking massive leads in medical education, research assistance and administrative workflow, LLMs have captured the attention of the medical fraternity. 1 However, away from the spotlight, a potentially more transformative development is quietly unfolding in the background with the maturation of small language models (SLMs).Unlike their data-hungry and computation-heavy counterparts, SLMs are designed for efficiency, explainability, and targeted deployment.In anaesthesiology, a speciality defined by time sensitivity, data accuracy, and critical decision making, SLMs may prove more clinically relevant than general-purpose LLMs.While LLMs have demonstrated their advantage across multiple aspects in anaesthesiology, their utility in real-time perioperative care remains constrained by infrastructure demands, latency, privacy concerns, and limited transparency.2 The question confronting clinicians today is not whether AI will enter anaesthesia practice, but whether it will do so in a form that is practical, safe, and fit for clinical purpose.SLMs may offer a compelling answer to this question.