Loading…
Glossary term
A compact language model designed to run with less compute, often on a laptop or phone.
Small language models have far fewer parameters, from a few hundred million to a few billion. They are cheaper and faster, and some run on a phone or laptop.
Well-trained small models, often distilled from larger ones, can match bigger models on narrow tasks.
They bring AI to cheap hardware, improve privacy by running locally and cut serving costs.