- input
- $0.00
- output
- $0.00
- Context
- 32Ktokens
- Released
- 1mo ago
Cost/1M tokens
Rerank-3 is a text reranking model from Voyage AI, part of MongoDB, that starts the Rerank 3 series alongside rerank-3-lite. In its launch announcement, MongoDB presents it as part of context engineering for agents, claiming improved accuracy across all tasks with the largest gains on long documents and code, and stating that rerank-3 outperforms Cohere Rerank v4.0 Pro by 2.72% overall. The model is built to rescore and reorder candidates from first-stage retrieval such as keyword or vector search, improving the quality and ordering of context delivered to agents.
How it compares
Frequently asked questions
API and pricing
Yes. VoyageAI by MongoDB lists a free endpoint with $0 input and output.
Yes, it looks reliable. In the last day, 1 endpoint had uptime data and 1 was at 99% uptime or better, with the best uptime at 100.0% by VoyageAI by MongoDB.
Capabilities and specs
The context window is 32K tokens (32,000). Maximum output is 29K tokens (28,800).
No. Its only listed input modality is text.
Quality and benchmarks
Rerank 5 Pro by Cohere, Rerank 5 Fast by Cohere, and Nemotron 3.5 Rerank 8B by NVIDIA. They were matched on model kind and release date, with no comparable price.
Rerank 5 Prorerank-3 vs Rerank 5 ProRerank 5 Fastrerank-3 vs Rerank 5 FastNemotron 3.5 Rerank 8Brerank-3 vs Nemotron 3.5 Rerank 8B
Basics
rerank-3 was made by VoyageAI by MongoDB and released on September 30, 2026.
It is unknown whether rerank-3 is open, but there are free API endpoints from VoyageAI by MongoDB.
Generated from Vertical Tabs data, updated November 9, 2026.
Discussion
No comments yet
Be the first to share your thoughts.