A local model is downloaded or deployed into a user-controlled environment such as a personal computer, workstation, private server, or local cluster. A compatible runtime loads the model and performs inference without requiring every prompt to reach a remote provider.
Local control can improve privacy, offline availability, customization, and cost predictability, but it transfers operational work to the user. Hardware capacity, model licensing, security updates, storage, and performance tuning become local responsibilities.
ELI5
A local model is an AI model stored and run on hardware controlled by the user or organization. Prompts do not need to be sent to a hosted model provider for every response.
For example, a company can run a document model on a private server so sensitive files stay inside its environment. The company then becomes responsible for hardware, security updates, model licenses, storage, and performance.
