In plain English
The biggest AI models need huge data centers. Smaller ones trade some broad knowledge for speed, low cost and privacy, and many can run right on your device, even without an internet connection.
Example
A note-taking app uses a small language model on your phone to summarize meetings, so the recording never gets uploaded.
In short
Big enough for the job. A lighter AI model that's fast, cheap and can run on your device.
Why it matters
Most everyday tasks don't need the largest model. Small models cut cost and latency, keep data on the device and power many AI features built into phones and laptops.
Common confusion
Small language model vs. LLM. There's no official size cutoff. "Small" is relative, and today's small models can beat the giant models of a few years ago. The real difference is where it runs and what it costs.