Cohere's Rerank 3 Nimble supports fast and accurate enterprise search applications
Cohere has launched Rerank 3 Nimble, a new AI model that enhances enterprise search and RAG systems by offering 3x faster performance, multilingual support, and improved accuracy across various tasks, available on Amazon SageMaker and for on-premise deployments.
Cohere's latest addition to its model lineup, Rerank 3 Nimble, is three times faster than its predecessor, Rerank 3, without a significant loss of accuracy. The model is available in two versions: English and multilingual. Rerank 3 Nimble can be deployed standalone to cut latency in time-critical search applications, such as retail, where website dropoff increases as response time lengthens. It can also be combined with Cohere's Command-R to build retrieval-augmented generation (RAG) applications. Rerank 3 Nimble can optimize any application by analyzing a query and a provided set of documents, passing on to Command-R only those that it finds most relevant for the query.
Rerank 3 Nimble delivers several benefits upon deployment, including enhancing search systems that leverage traditional search algorithms, performing accurate searches in complex documentation, enabling multilingual support in generative AI applications, and figuring out the most relevant information for a RAG system from different data stores. Cohere's customers reap the benefits of Rerank in their applications: IT service management company Atomicwork saw a +20% improvement in the accuracy of Atom AI, the company's RAG-enabled AI assistant.
The Rerank 3 Nimble official launch announcement showcases the model's remarkable performance in enterprise tasks-related benchmarks. The model is available on Amazon SageMaker and for on-premise deployments only, and it will be accessible on Amazon Jumpstart starting July 30, 2024.
