GLOSSARY
Small Language Model (SLM)
Compact models (roughly 1–13B parameters) tuned to punch above their weight — fast, cheap and private enough to run on phones, laptops and edge devices.
The scaling laws rewarded bigness, so for years small meant mediocre. Modern SLMs changed that: curated training data, distillation from larger teachers and better recipes give 3–8B models usable general ability plus surprising depth in narrow domains. They are the models that make local AI (Ollama), on-device assistants and cheap API tiers practical.
The decision rule is refreshingly simple: use the smallest model that passes your evals. Small means lower latency, lower cost, easier privacy compliance and the option to run on-device — and the evals, not folklore, decide where the line sits for your task.