SantageAI Glossary › Foundation Model
AI Glossary

What is a Foundation Model?

A foundation model is a large AI model trained on broad data at scale that can be adapted to a wide range of downstream tasks.

What is the core idea behind foundation models?

A foundation model is a general-purpose base you adapt, not a single-purpose tool you build from scratch.

How do foundation models differ from related concepts?

ConceptDifference
Foundation Model vs LLMLarge language models are foundation models for text. Foundation models also include image, audio, and multimodal models.
Foundation Model vs Fine-Tuned ModelA foundation model is the general base. A fine-tuned model is that base adapted to a specific task.
Foundation Model vs Narrow AINarrow AI is built for one task. A foundation model is built to be adapted to many.

How do foundation models work?

What are examples of foundation models?

Why are foundation models important?

Foundation models shifted AI from building a new model per task to adapting one powerful base model to many. This is the economic and technical engine behind the current AI boom. The term was coined by Stanford researchers in 2021.

How are foundation models used in practice?

Foundation models underpin chatbots, coding assistants, image generators, and enterprise AI platforms. Concerns include concentration of power among a few labs, cost, bias, and the difficulty of auditing what they have learned.

Frequently Asked Questions

Is a foundation model the same as a large language model?
Not exactly. All large language models are foundation models, but foundation models also include image, audio, and multimodal systems, not just text.
Who makes foundation models?
Mainly large AI labs such as OpenAI, Anthropic, Google, Meta, and DeepSeek, because training them requires enormous data and compute.