rerank-2
Rerank 2 expands Voyage's earlier reranking family with improved general and multilingual retrieval.
Model design
The release evaluates 93 datasets, including 51 multilingual datasets in 31 languages. Its 16,000-token pair limit is twice that of rerank-2-lite. Both models score a query and document jointly.rerank-2 and rerank-2-lite
Practical use
Practical guidance. Use the larger pair limit when evidence is spread across a long candidate document. Keep an eye on which text reaches the model; silent truncation can remove the passage needed to judge relevance.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API |
| Context window | 16,000 tokens Query and one document together. Voyage AI reranking API and earlier models |
| Atlas lifecycle | Legacy since 11 August 2025MongoDB Voyage AI model lifecycle: Atlas scope and release history |
| Query limit | 4,000 tokensVoyage AI reranking API and earlier models |
Related models
- rerank-2.5
Voyage AI · Voyage Rerank 2.5
Reranker that follows natural-language ranking criteria.
Launch date API - rerank-2-lite
Voyage AI · Voyage Rerank 2
Smaller second-generation reranker with an 8,000-token pair limit.
Launch date API Earlier general and multilingual relevance reranker.
Launch date API