What Voyage AI does
Voyage AI develops embedding models and rerankers designed to maximize retrieval accuracy for RAG systems and semantic search over unstructured data. Its catalog spans general-purpose embeddings (voyage-3.5 and lite variants), multimodal embeddings, and domain-specific models tuned for code, finance, and legal text, plus rerankers (rerank-2.5 and lite) that improve relevance and follow instructions. Highlights include long context windows, much shorter output vectors for cheaper storage, and a Batch API for large-scale workloads.
Research roots and acquisition
Founded in September 2023 by Stanford professor Tengyu Ma with a team of researchers from Stanford, MIT, Berkeley, and Princeton, Voyage AI quickly became a go-to provider of high-quality retrieval models. In February 2025, MongoDB acquired Voyage AI for approximately $220M to strengthen AI accuracy across its database products. Voyage continues to operate under its own brand with a dashboard, API, and documentation, and its models are also being integrated into MongoDB's platform, making it a key embeddings provider for AI builders.