# What is CLIP? Definition and meaning

A model trained contrastively on image and caption pairs that embeds both into one space.

## Definition

**CLIP**: A model trained contrastively on image and caption pairs that embeds both into one space.

In short: Joint image and text embeddings.

## When it is used

Zero-shot image classification and cross-modal search.

## Learn more

The full guide to **[CLIP](/tools/clip)** explains how it works, when to use it and how it compares with the alternatives, with examples. The [glossary](/glossary) lists every term on the site from A to Z.

Previous term: [ClickHouse](/glossary/clickhouse) · Next term: [Cloud TPU](/glossary/cloud-tpu)
