In plain words
In machine learning, a model is a mathematical or computational system used to make predictions or produce outputs. Its behavior depends on its architecture and the parameters learned during training.
A closer look
A model may be simple, such as a linear formula, or large, such as a neural network with billions of parameters. The architecture specifies how calculations are arranged. Training determines values for adjustable parameters. Inference runs those calculations on a new input.
People also use “model” to mean a particular trained checkpoint, a named product version, or a family of related systems. These are different levels of description. A chat application is usually more than its model: it may include instructions, search, tool access, a user interface, and stored information.
In practice
A model estimates a home’s price from its size and location. Another model generates a paragraph from a prompt. Both map inputs to outputs using a learned computational structure.
A useful distinction
A model is not simply a searchable folder of its training documents. Training can encode patterns and sometimes memorized material in parameters, but the model’s ordinary output process is computation, not document lookup.