A language model with fewer parameters (typically \<10B) designed for efficiency, lower latency, and often local/edge deployment. Unlike larger models, SLMs can often run on consumer hardware (laptops, phones) without sending data to a cloud provider, changing the privacy and security risk profile.