Cohere Rerank is a model, offered through Cohere's API, that takes a search query plus a set of candidate texts and sorts those texts so the best matches come first.
In plain words
A first search step, such as keyword search or vector search, returns a list of candidates. A reranker reads the query and each candidate document together, scores how well they match, and sorts them. So people usually run Rerank on the results of an existing search. Recent versions handle many languages and semi-structured data such as JSON.
A quick example
Imagine a help-centre search that returns 50 articles for "reset my password on the app". Sending the query and those 50 articles to Rerank puts the article that actually explains the app's reset steps at the top.
Where you'll meet it
- Improving search results before showing them or passing them to an LLM.
- The reranking step in RAG pipelines.
Related
Sources
- DocsAn Overview of Cohere's Rerank Model, Cohere Docs · read 29 Sept 2026
- DocsCohere's Rerank model, Cohere Docs · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.