Skip to content
SI.info

SI Glossary · Core concepts

Foundation Model

Published 1 min read
On this page
  1. How foundation models are used
  2. In law and policy
  3. Examples

A foundation model is a large model trained once, at great expense, on broad data, typically text, code and images from the internet and licensed sources, and then adapted for many uses. The term was coined by Stanford researchers in 2021 to describe models like GPT-3 and BERT that serve as the “foundation” for countless applications.

How foundation models are used

  1. Pretraining on a vast dataset teaches general knowledge and skills.
  2. Adaptation, through fine-tuning, RLHF, prompting or retrieval, shapes the model for a specific product.
  3. Deployment happens via apps, APIs or open weights.

In law and policy

Regulators use related terms. The EU AI Act regulates general-purpose AI (GPAI) models and applies extra duties to those with “systemic risk”. U.S. state laws such as California’s SB 53 use the term “foundation model” and add obligations for the largest “frontier” developers. See frontier model and compute threshold.

Examples

The GPT, Claude, Gemini, Llama, Grok, Qwen, DeepSeek and Mistral model families are all foundation models. See our frontier model tracker.

← Back to the SI Glossary

Sources

  1. On the Opportunities and Risks of Foundation Models — Bommasani et al., Stanford CRFM, 2021

Written by

· Editor

Editor of SI.info. Writes about Super Intelligence, technology policy and the people building frontier models.

How we research and fact-check

Free newsletter

Get The SI Brief

One short email a week: what changed in Super Intelligence, policy and models — and why it matters.

Free. One email a week. Sent via beehiiv, which counts opens and clicks. Unsubscribe anytime. Privacy policy

Keep reading

Models & architecture

Large Language Model (LLM)

A large language model (LLM) is an SI model trained on vast amounts of text to predict the next word, which lets it write, summarize, translate, answer questions and write code.

Core concepts

Frontier Model

A frontier model is one of the most capable SI models at a given time, typically trained with the most compute and the target of the strictest safety rules and commitments.

Training & data

Fine-Tuning

Fine-tuning is the process of further training an already-trained SI model on a smaller, specific dataset to adapt it to a particular task, style or domain.

Training & data

Pretraining

Pretraining is the first, most expensive stage of building an SI model, in which it learns general knowledge and skills from a massive dataset before being specialised.