What it is:

The Voyage 4 series is a new generation of text embedding models consisting of voyage-4-large, voyage-4, voyage-4-lite, and the open-weights voyage-4-nano. All models in the series share a compatible embedding space, eliminating the constraint of using the same embedding model at both document indexing and query time. Users can index with voyage-4-large for maximum retrieval quality, then query with any Voyage 4 model to optimize quality-latency-cost tradeoffs per use case.

Who it’s for:


These new models are designed to serve teams seeking even more accurate retrieval, as well as the new wave of AI developers building context-engineered agents and AI systems with long-term memory.

Why it matters:


Previous generations of embedding models required using identical models to embed both queries and documents. By sharing embedding spaces between models, the Voyage 4 model series enables flexibility in the way embeddings are generated: for example, using voyage-4-large for document/chunk embeddings and voyage-4-lite for query embeddings.

How to get started:

Sign up, generate a model API key, and get 200M free tokens on our latest models. Dive into the documentation and start building with the quick start.