What is Small Language Models (SLMs)

Definition

Small language models (SLMs) are compact language models with far fewer parameters than large LLMs, designed to run efficiently on limited hardware or at lower cost while still handling many language tasks well, often after being specialised for a domain.

« Back to Glossary Index

Run efficiently on modest hardware, including edge devices
Cost far less to deploy and operate than large LLMs
Offer lower latency for responsive applications
Can match large models on focused, specialised tasks

Real World Example

A company replaces a large model with a fine-tuned small language model for its narrow classification task, cutting cost and latency dramatically while maintaining accuracy on that specific job.

FAQs

How do SLMs differ from LLMs?

SLMs have far fewer parameters, making them cheaper, faster, and able to run on limited hardware, at some cost to broad capability.

When are SLMs preferable?

For focused tasks, cost-sensitive or low-latency uses, on-device deployment, or privacy-driven local inference.

Can SLMs match LLMs?

On narrow, well-defined tasks, especially after fine-tuning, SLMs can rival much larger models.