Voyage AI

High-Performance Embeddings for RAG

Voyage AI provides state-of-the-art embedding models specifically optimized for retrieval-augmented generation applications, offering best-in-class retrieval accuracy that enables RAG systems to find more relevant context and generate better AI answers. Founded by AI researchers from Stanford and other leading institutions, Voyage focuses exclusively on building the best embedding models for search and retrieval tasks.

RAG-Optimized Architecture

Voyage's models are not general-purpose embeddings repurposed for retrieval — they are designed from the ground up for the specific task of finding semantically relevant documents when answering questions. The models excel at the asymmetric retrieval task where a short query is matched against longer document passages, which is exactly how RAG systems operate. This specialization translates directly into higher RAG answer quality.

• Domain-specific embedding models for code, finance, law, and general use
• Optimized for asymmetric retrieval in RAG applications
• Reranking models for two-stage retrieval pipelines
• Multimodal embeddings supporting text and images
• Context-aware embeddings for document chunking

For RAG-Focused Developers

Developers and teams building RAG applications where retrieval quality directly impacts answer quality use Voyage to squeeze maximum performance out of their retrieval pipeline. The combination of top embedding quality and specialized reranking models creates retrieval pipelines that consistently outperform alternatives on real-world RAG benchmarks.