Data as of Oct 8, 2026Based on 33,992 AI responses
Reviewed by Dimitry Apollonsky ·
Voyage AI provides reranker models that take a user query and a set of candidate documents (typically retrieved by embedding-based or lexical search) and produce refined relevance rankings using cross-encoder architectures. The service offers multiple models (rerank-2.5, rerank-2, rerank-1 and their lite variants) with varying context lengths and latency to balance quality and performance, including multilingual support. Access is via a Python API (the voyageai package) using voyageai.Client.rerank to rerank up to a configurable number of documents with token limits after an initial retrieval step.
Products
<1%No change
of AI answers about rerank-3-lite and its rivals. Since Jul 5
The market map
RAG Platforms and Retrieval ToolsSince Jul 5
rerank-3-lite's share in each topic, as its page ranks it
Rank
Mentioned in
Where rerank-3-lite ranks in AI
