Small Language Model (SLM)
A compact language model — typically under ~15 billion parameters — designed to run efficiently on local or edge hardware while remaining capable for focused tasks.
What Is Small Language Model (SLM)?
A Small Language Model (SLM) is a language model with a relatively low parameter count — commonly from a few hundred million up to around 15 billion parameters — built to deliver strong performance on focused tasks while running cheaply and quickly. SLMs trade some of the broad, general capability of frontier models for efficiency: lower latency, lower cost, smaller memory footprint, and the ability to run on a laptop, phone, or edge device without cloud access.
Modern SLMs such as Microsoft's Phi series, Google's Gemma, Meta's smaller Llama models, and Qwen's compact variants punch well above their weight thanks to high-quality, carefully curated training data and distillation from larger models. For many production tasks — classification, extraction, routing, summarisation of short text — a well-chosen SLM matches a frontier model at a fraction of the cost and can be fine-tuned for a specific domain on modest hardware.
Why It Matters
Small Language Models are central to making AI practical and private at scale. They enable on-device AI (no data leaves the user's hardware), drastically lower inference costs for high-volume workloads, and support offline use cases. As organisations move from experimentation to production, the question is increasingly "what is the smallest model that reliably does this task?" rather than "what is the most powerful model available?"
Real-World Examples
On-Device Assistant
A phone runs a 3B-parameter SLM locally to summarize notifications and draft replies without sending data to the cloud.