Story details
Canadian AI company Cohere announced its Rerank 5 family on 21 October. Rerank Pro is the quality-focused option for complex, reasoning-intensive retrieval and fields such as finance, healthcare and scientific research, while Rerank Fast is designed for low latency and high query volumes. Cohere says Fast outperforms its previous Rerank 4 Pro model.
Both models support more than 100 languages, a 32,000-token context window and ranking of semi-structured records formatted as JSON or YAML. The optional Context Focus feature can identify relevant passages within documents in the same call; Cohere says this can reduce the text sent onward to a language model. The models are available through Cohere's API, Microsoft Foundry and Amazon SageMaker, with single-tenant deployment through Model Vault.

