Cohere releases the Embed 5 family of embedding models

On October 2, 2026, reports said Cohere released Embed 5, a new family of embedding models. The family is positioned for three kinds of work: enterprise search, RAG, and agentic retrieval. It ships in two tiers.

Embed 5 Pro targets maximum retrieval quality, while Embed 5 Fast targets latency and cost on the live query path. Both tiers accept text, images, and fused text-plus-image inputs, so the same family covers pure text retrieval as well as multimodal and mixed-modality retrieval.

The report did not include parameter counts, context length, pricing, or benchmark scores, and it did not explain how Embed 5 differs from earlier Cohere embedding models. A truncated headline fragment mentions a comparison with Voyage 4 Large, but the visible content cuts off at that point, so the comparison itself is not developed and no results are given. The report also does not state whether the models are generally available or how they can be accessed. As of this single report, the release, the positioning of the two tiers, and the supported input types are all that can be verified.