**VoyageAI Unveils Multimodal Embedding Model: "Voyage-Multimodal-3"**

A new multimodal embedding model, "Voyage-Multimodal-3," has been introduced by VoyageAI. This model allows for efficient vectorization of knowledge bases containing both pure-text documents and unstructured data (such as PDFs/slides/webpages). The model's performance is evaluated across 20 multimodal datasets, demonstrating its effectiveness in tasks like table/figure retrieval, document screenshot retrieval, and text-to-photo retrieval.

The model outperforms several top-performing alternatives in these tasks, including OpenAI CLIP large and Cohere multimodal v3. Voyage-Multimodal-3 is now available for use, with the first 200 million tokens free.

Source: https://blog.voyageai.com/2024/11/12/voyage-multimodal-3/