A large language model is a transformer-based neural network that predicts text and, through that ability, can summarize, translate, answer questions and produce structured output such as JSON. For IoT, LLMs are mostly used as an interface layer: turning operator questions into queries over telemetry, explaining alarms in plain language, or driving an AI agent that orchestrates devices.
LLMs are far too large for most edge hardware, so they usually run in the cloud or on a local server, with the devices themselves exposing APIs. Smaller distilled models can run on a gateway for limited tasks. The trade-offs are cost per query, latency, and the fact that model outputs are probabilistic and must be validated before they touch actuators. See AI agents on the factory floor.
Part of the IoT Glossary. Spotted an error or a missing term? Tell us.