What is a foundation model?

A foundation model is a large, general-purpose AI model that can be adapted to many different tasks. It is trained on broad data, such as much of the public internet, books and code. GPT, Claude, Gemini and Llama are foundation models. Businesses and software companies build products on top of them instead of building their own models from scratch.

Why it matters for a business

Almost every AI product a business buys is a foundation model from one of a few labs, adapted with prompting, retrieval and sometimes fine-tuning. “We’ve built our own model” usually means exactly that: a layer on top of someone else’s foundation model, because training one from scratch costs far more than any single business problem justifies. Which foundation model sits underneath decides whose data terms apply, and whether the product can switch when a better model ships.

Go deeper

AI vs. machine learning, and where LLMs fit

Related terms

• Large language model (LLM)
• Frontier model
• Fine-tuning
• AI wrapper
• Open-weight model

All terms, A to Z