Cohere Rerank

About Cohere Rerank

"Cross-encoder reranking API to dramatically improve RAG and search relevance"

Cohere Rerank is a powerful relevance reranking API that dramatically improves search and RAG quality by using a cross-encoder model to score the true relevance of retrieved documents to a query. Unlike embedding-based retrieval that uses vector similarity, Rerank understands the nuanced relationship between queries and documents, filtering out irrelevant results and surfacing the most useful information. Adding Rerank as a post-processing step to any retrieval pipeline — including keyword search, vector search, or hybrid search — consistently boosts answer quality with minimal code changes.

• Cross-encoder relevance scoring
• Works with any retrieval system
• Multi-language support
• Low-latency API
• Measurable accuracy improvement

Official Links