Foundation Model

A foundation model is a large AI model trained on broad, unlabeled data that can be adapted to a wide range of downstream tasks. Large language models are the best-known type of foundation model, but the category also includes image, audio, and multimodal models.

How it works

The term (coined by Stanford in 2021) captures a shift in AI: instead of training a separate model for each task, organizations train one general-purpose model on massive data, then adapt it — via fine-tuning, prompting, or retrieval-augmented generation — to many specific uses. Examples include GPT, Claude, Gemini, and Llama.

Thank you for reading this post, don't forget to subscribe!

Why it matters

Foundation models concentrated AI development: nearly 90% of notable AI models in 2024 came from industry. The capital and compute required to train them is a major driver of AI investment — corporate AI investment hit $252.3 billion in 2024. See Generative AI Statistics 2026.