voyage-multimodal-3.5: a new multimodal retrieval frontier with video support
Summary
Voyage AI announces voyage-multimodal-3.5, a next-gen multimodal embedding model that unifies text, images, and video in a single embedding space for retrieval. The model adds explicit video support, Matryoshka embeddings with configurable dimensions, and quantization options, delivering higher retrieval accuracy on visual document and video tasks while staying competitive with text-only baselines.