Introducing the Voyage 3 series embedding models, voyage-3 and voyage-3-lite, pushing boundaries in retrieval quality, latency, and cost efficiency. voyage-3 outperforms OpenAI v3 large by 7.55% across various domains with lower costs and smaller dimensions, enhancing vectorDB affordability. voyage-3-lite offers superior accuracy over OpenAI v3 large while being more cost-effective. Both models support a 32K-token context length, surpassing competitors. The Voyage 2 series also excels in domain-specific models, like voyage-multilingual-2, for superior multilingual performance. With advanced research innovations and alignment with human feedback, upgrading to Voyage 3 promises enhanced retrieval quality at a lower cost.
https://blog.voyageai.com/2024/09/18/voyage-3/