A foundation model is a large AI model trained on massive amounts of data that can be adapted for a wide range of tasks. Rather than being built for one specific purpose, it serves as a starting point — a foundation — that other products and applications can build on. Most modern large language models, including GPT-4 and Claude Sonnet, are foundation models.
Foundation models are made available to other organizations primarily through APIs, meaning a company can build a product powered by GPT-4 or Claude Sonnet without training their own model from scratch. The model provider handles the hard part; the product team builds the experience on top of it.
Why This Matters for Technical Writers
If you’re documenting an AI-enabled product, there’s a good chance it’s built on a foundation model rather than a proprietary one. This means the behavior you’re documenting is shaped by two things: the underlying model and the decisions the product team made in how they deployed it. Some capabilities and limitations come from the foundation model itself; others come from the product layer on top. Knowing which is which helps you write more accurate documentation and ask better questions of your engineering team.
Common Confusion
Foundation model and LLM are often used interchangeably, but LLMs are a subset of foundation models. Not all foundation models work with language. There are foundation models for images, audio, and video as well.
Related Terms
large language model (LLM), AI model, generative AI, fine-tuning, API