The company introduced a new suite of retrieval tools designed for corporate data across more than 100 languages. This Embed 5 family includes a Pro variant that beats Gemini Embedding 2, Voyage 4 Large, and Jina Embeddings v5 in average scores on the ViDoRe V3 benchmark, and a Fast version for low-latency performance. The models also support the retrieval of images and parsed PDF documents for wider enterprise utility.

The tools are available through the Cohere API, Model Vault, North, Microsoft Foundry, and Amazon SageMaker. Because the Pro and Fast models share a single embedding space, users can index their information with one version and retrieve it using the other without a loss in performance.