Evidence-first authority page
What Is a Foundation Model?
A definition of foundation models, why the term is broader than 'large language model,' and how they relate to fine-tuned derivatives.
Direct answer
A foundation model is a large model trained on broad, diverse data that serves as a base for many downstream applications, either used directly or adapted through fine-tuning — the term is broader than 'large language model' since it also covers vision, audio, and multimodal base models trained the same way.
Last reviewed 2026-08-29
This page remains noindex while its page-level source receipts are completed and verified. It is available for review, not presented as publication-ready authority.
Why 'foundation' rather than just 'large model'
The term emphasizes the model's role as a base other systems build on, rather than its size specifically — a foundation model is characterized by broad, general training that transfers to many tasks, not by parameter count alone. A smaller model trained the same general-purpose way is still a foundation model; a large model trained narrowly for one task typically isn't described that way.
Foundation model vs. fine-tuned derivative
A foundation model is often the starting point; a fine-tuned derivative is that same base model further trained on a narrower dataset for a specific task or domain. The foundation model provides general capability; fine-tuning specializes it, typically at the cost of some generality outside the fine-tuned domain.
Why this distinction matters when evaluating a product
A vendor's claim about 'our model' can mean a foundation model used as-is, a fine-tuned version of someone else's foundation model, or a foundation model they trained themselves — these are materially different claims about the underlying technology and worth distinguishing when evaluating a product's actual capability and differentiation.