What is an AI model?

An AI model is the trained core of an AI system: a program that has learned from data to turn an input into an output. For example, it turns a question into an answer, or a photo into a description. It is built by training on large numbers of examples, and what it learns is stored as a very large set of numbers called weights. Apps such as ChatGPT are built around a model and add the interface, accounts and other features.

Why it matters for a business

A model on its own has no interface, no memory and no access to anything; it turns input into output. ChatGPT is a product built around OpenAI’s models, and the same model can sit behind many different products, so the model is only part of what decides how good an AI tool is. Each model is also a snapshot of one training run. A new version can behave differently under the same name, which is why systems that matter call a fixed, dated version.

Go deeper

What an AI model actually is, and how LLMs work

Related terms

• Large language model (LLM)
• Foundation model
• Training data
• Inference
• Open-weight model

All terms, A to Z