voyage-multimodal-3.5
Voyage Multimodal 3.5 adds video input to Voyage's text-and-image retrieval family. It supports variable output dimensions.
Model design
The model can represent mixed text, screenshots, slides, tables and video frames in a shared space. Its supported output sizes are 256, 512, 1,024 and 2,048 dimensions. It retrieves media rather than generating a written account of them.voyage-multimodal-3.5
Practical use
Practical guidance. Keep page numbers, crop locations or frame times alongside each vector. A relevant frame needs a usable link back to the video. Evaluate visual queries separately from text-only queries so a large text test set does not hide weak image retrieval.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API |
| Atlas lifecycle | ActiveMongoDB Voyage AI model lifecycle: Atlas scope and release history |
| Input limit | 32,000 tokens, including media tokensVoyage AI multimodal embedding API |
Related models
General text embeddings compatible with the other Voyage 4 models.
Launch date API- voyage-multimodal-3
Voyage AI · Voyage Multimodal
Joint text and image embeddings for visual document retrieval.
Launch date API