Reranking Models
AI models for reranking search and retrieval results
Rerank
All model series
3 seriesChoose a reranker for search results
Reranking scores candidate documents against a query after retrieval. Measure whether it improves the ordering of your actual search results enough to justify added latency and cost.
Selection signals
- Supply a representative candidate set; a reranker cannot recover relevant documents that retrieval omitted.
- Verify query length, document limits, and the supported number of results.
- Compare relevance and end-to-end latency with and without reranking on the same judged queries.