# Foundation Model

> A foundation model is a model trained on broad data at scale, usually with self-supervision, that can be adapted to a wide range of downstream tasks by prompting or fine-tuning.

- Identifier: PTL-0103
- Category: Foundations
- Canonical URL: https://protologue.com/t/foundation-model/
- Also known as: frontier model, general-purpose model
- Introduced: 2021

## Description

The term was introduced in 2021 by researchers at Stanford to describe models such as BERT, GPT-3, and CLIP that serve as a common base for many applications. Large language models are the most prominent kind, but the term also covers image, audio, and multimodal models. Because many systems are built on the same few foundation models, their strengths and flaws are inherited widely.

## Narrower terms

- [Large Language Model](https://protologue.com/t/large-language-model/)

## Related terms

- [Base Model](https://protologue.com/t/base-model/)
- [Fine-Tuning](https://protologue.com/t/fine-tuning/)

## Sources

- Bommasani et al. (2021). On the Opportunities and Risks of Foundation Models. https://arxiv.org/abs/2108.07258

## Cite this entry

Protologue. (2026). Foundation Model. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0103). https://protologue.com/t/foundation-model/

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)