Small Language Model (SLM) in plain English.

A small language model (SLM) is a compact AI model — typically under 10 billion parameters, sometimes far smaller — designed to run quickly and cheaply, often on a phone, laptop, or modest server rather than a data center full of GPUs. Examples include Microsoft's Phi series, Google's Gemma, and Llama 4's smaller variants. SLMs trail frontier models like GPT-5.6 or Claude Opus 4.8 on broad, open-ended tasks, but for narrow, well-defined jobs — classifying support tickets, extracting fields from a form, running on-device autocomplete — a well-tuned SLM can match a much larger model at a fraction of the cost and latency. The 2025–2026 trend has been "use the smallest model that reliably does the job," with frontier models reserved for genuinely hard reasoning and SLMs handling high-volume, narrow tasks in the background of most production AI systems.