# Small Language Models (SLMs) Language models with relatively few parameters (typically under 10B) designed for efficiency, on-device deployment, and domain-specific tasks. Examples: Phi (Microsoft), Gemma (Google), Llama 3.2 1B/3B (Meta). SLMs challenge the assumption that bigger is always better; with good training data and distillation, small models achieve surprisingly strong performance on targeted tasks. Key advantages: low latency, low cost, privacy (runs locally), deployable on mobile/edge devices. Trade-off: less general capability than large models; excel at specific tasks, struggle with broad reasoning. ## References ## Related - [[Large Language Models (LLMs)]] - [[Knowledge Distillation]] - [[Gemini Nano]] - [[On-Device Machine Learning]] - [[Edge AI]] - [[Browser-Provided Language Models]] - [[Transformers.js]]