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.