Cohere offers the Embed model family, which converts text and images into numerical representations called embeddings so that enterprise search systems can find documents by meaning rather than exact keyword matches. The most recent version, Embed 4, supports over 100 languages and can process both text and images together.

The Embed models (currently Embed 4, released in April 2025) take text or image inputs and produce vector embeddings, which are numerical codes that capture semantic meaning. These embeddings power search systems, classification, clustering, and RAG pipelines. Embed 4 is available on the Cohere Platform, AWS SageMaker, and Azure AI Foundry, and supports multilingual and multimodal enterprise search including PDFs with mixed text and images.