All-in-one embedding model for interleaved text, images, and screenshots

Voyage AI introduces voyage-multimodal-3, a new all-in-one embedding model designed for interleaved text, images, and screenshots, offering a major advancement for multimodal embeddings and semantic search. It features significant improvements over previous models in vectorizing visual texts and complex document features, resulting in a 19.63% increase in retrieval accuracy.

• Launch date: November 12, 2024
• 19.63% accuracy improvement
• Handles interleaved data
• Vectorizes mixed modalities
• Eliminates complex parsing