AI Model

An AI model, often just "model", is the artifact produced by training: a structure of numbers (parameters or weights) that maps inputs to outputs. It is separate from the code that runs it and from the data that shaped it.

Why it matters

Choosing a model means trading accuracy, speed, cost, and openness. Providers release versions that behave differently, so pinning a version and testing before switching protects your application from surprising changes.

Example

A model is typically described by its type, size, and version:

text-classifier-v3   (~110 million parameters)

Saved to disk it is a file of weights plus a config. Loading it and feeding it input is inference. Larger models usually need more memory, often a GPU.

Common misconceptions

• The model and the product are the same thing. A product adds prompts, data, and safeguards around a model.
• Newer versions are always better on your task. Test them on your own cases.
• A model contains a searchable copy of its training data. It stores learned patterns, not a database.

Related

• Neural network
• Training data
• Benchmark
• Machine learning

This page is general information written by our editorial team and updated when documentation changes. Servers and providers differ, so test changes safely first. See our editorial policy and disclaimer.