What is a foundation model?
A foundation model is a large AI model trained on a very broad set of data so it can be adapted to many different tasks. Large language models are the best-known example, and the same idea applies to image, audio, and multimodal models.
Last reviewed .
Why they're called foundation models
Instead of building a separate model for every task, developers start with one general-purpose model and build on top of it. That shared base is the foundation.
Adapting it can be as simple as writing a good prompt, or as involved as fine-tuning it on specialized examples.
Foundation models and LLMs
A large language model is a type of foundation model that focuses on text. Other foundation models work with images, audio, video, or several of these at once.
Generative AI tools such as writing assistants and image generators are commonly built on foundation models.
Choosing one
Models differ in quality, speed, cost, and the data they were trained on. Test candidates on your own tasks, and check each provider's terms for how your data is handled.
Common questions
Is a foundation model the same as an LLM?
Not exactly. An LLM is one kind of foundation model, focused on language. Foundation models also include those built for images, audio, and other data.
Who builds foundation models?
They're usually built by large technology companies and research labs, because training them takes large amounts of data and computing power. Most businesses use them through a service or by adapting them.
Can you customize a foundation model?
Yes. You can guide it with prompts, supply documents through retrieval, or fine-tune it on your own examples.
Do foundation models have biases?
They can, because they learn from large datasets that reflect human-created content. Testing outputs and keeping people involved in review helps catch problems.
Want help putting this to work?
Tell us where your marketing stands today, and we'll show you what to do first.