Text embedding models are neural networks that transform texts into numerical vectors. They are a crucial building block for semantic search/retrieval systems and retrieval-augmented generation (RAG) and are responsible for the retrieval quality. voyage-3-lite is a lightweight general-purpose embedding model optimized for latency and cost, which: [1] outperforms OpenAI v3 large and small by 3.82% and 7.58% on average across the domains, respectively, [2] has a 6-8x smaller embedding dimension (512) compared to OpenAI (3072) and E5 Mistral (4096), resulting in 6-8x lower vectorDB costs, and [3] supports a 32K-token context length, compared to OpenAI (8K) and Cohere (512). Learn more about voyage-3-lite here.