Model release · Cohere
Rerank 5
- Model released
- Review updated
Cohere says Rerank 5 Pro outperformed every model it tested on benchmarks including nanoMTEB, ViDoRe V3 and BRIGHT. Fast targets high-volume, latency-sensitive workloads, with nearly three times Pro’s throughput. Context Focus returns relevant passages and scores; Cohere reports token savings of up to 41%. Both variants have a 32K-token context limit per document, including the query.
Availability
Pro and Fast are generally available through the Cohere API, Model Vault, Microsoft Foundry and Amazon SageMaker, and to all North users.