Cohere has released Embed 5, a new family of embedding models for enterprise search and retrieval. The launch was announced on October 1.

Embed 5 comes in two tiers. Embed 5 Pro is priced at $0.12 per million tokens and is optimised for maximum quality. Embed 5 Fast costs $0.08 per million tokens and targets latency-sensitive workloads.

Both tiers share a single embedding space. That means teams can build an index with Pro and query it with the cheaper Fast model without rebuilding anything.

The models support 128,000-token inputs, more than 100 languages, and multimodal text-plus-image inputs. Cohere reports that Pro scores 85.8 on the ViDoRe V3 retrieval benchmark, ahead of rivals from Voyage, Google and OpenAI.

Embed 5 is available on the Cohere API and Model Vault, Microsoft Foundry and Amazon SageMaker. It is aimed at search, retrieval-augmented generation and AI agent workflows.

Cohere will discuss Embed 5 and its Parse 5 document parser in a public session on X on October 8. The company has also put its Compass Cloud managed-search product into private beta for retrieval and agentic workloads.