Voyage AI

Voyage AI is part of MongoDB but its models remain directly available beyond MongoDB. This page covers Voyage embedding and reranking models, not Atlas Vector Search, vector storage, or general RAG architecture.¹³⁴

For: Teams prepared to benchmark retrieval-focused embeddings and rerankers against their own documents and queries

• Another provider wins corpus-specific evaluation
• The selected model family or distribution route does not meet data requirements
• Re-embedding and provider migration cost outweigh the relevance benefit

• Retrieval focusVoyage publishes embedding families for general, code, multilingual, contextual, and multimodal retrieval.¹³⁴
• RerankingSeparate reranker endpoints support a second relevance stage after candidate retrieval.¹³⁴
• Distribution choicesVoyage models remain directly available and are also distributed through selected cloud and MongoDB routes.¹³⁴

Usage pricing varies across text tokens, multimodal pixels, reranker tokens, Batch, Files storage, and fine-tuned or dedicated arrangements.³

Text models

Token-based by active model³

Multimodal

Text tokens plus image or video pixels³

Migration

Re-embedding, reranking, storage, and evaluation remain application costs³