Foundation model

A foundation model is trained broadly enough to be adapted or used across multiple downstream tasks and applications.

Plan for this page

Understand one concept in plain language and connect it to nearby ideas.

You will leave with: Its initial training creates reusable capabilities or representations rather than one final narrow product. Developers may adapt it through prompting, retrieval, tools, fine-tuning, or other layers, each of which changes system behavior and risk.
Time: 3–8 min
Before you begin: No prerequisite. Keep one real example in mind.
Do this now: Read the explanation, then compare the example with the common misunderstanding.

Prepared next stepConnect models to products

Foundation modelA foundation model is trained broadly enough to be adapted or used across multiple downstream tasks and applications.

Model
Large language model
Fine-tuning

Plain explanation

Its initial training creates reusable capabilities or representations rather than one final narrow product. Developers may adapt it through prompting, retrieval, tools, fine-tuning, or other layers, each of which changes system behavior and risk.

Example

One broadly trained language model can support summarization, classification, question answering, and drafting in several different products.

Why it matters

Foundation models concentrate capabilities and limitations that can propagate into many applications. Evaluating the base model does not replace evaluation of each adapted system and context.

Common misunderstanding

Misconception: a foundation model is universally capable or foundational truth. Correction: “foundation” refers to reuse across applications, not completeness, reliability, or authority.

Related concepts

Follow relationships, not only the alphabet

ModelA model is a learned or designed representation that maps inputs to outputs for a defined task or family of tasks.Large language modelA large language model, or LLM, is a parameter-rich model trained on language sequences to predict and generate tokens in context.Fine-tuningFine-tuning is additional training that adapts a pretrained model to selected tasks, domains, examples, or behavior goals.

Connect models to products

Sources and scope

What supports this object

Machine Learning GlossaryGoogle for Developers

Provides current technical vocabulary for foundation models; boundaries and naming vary across institutions and regulations.

Open source

Loading page content...

…