voyage-multimodal-3
Voyage Multimodal 3 is Voyage's first multimodal embedding release. It accepts text, images and interleaved combinations.
Model design
A page image can retain charts and layout that plain text extraction would lose. The API gives this version a fixed 1,024-dimensional output; explicit video input belongs to 3.5.Voyage AI multimodal embedding API
Practical use
Practical guidance. Use page images when figures or layout carry evidence that extracted text misses. Keep the source document and page index available after retrieval. Image similarity alone does not establish that a chart supports a proposed numerical claim.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API |
| Atlas lifecycle | Legacy since 15 January 2026MongoDB Voyage AI model lifecycle: Atlas scope and release history |
| Input limit | 32,000 tokens, including image tokensVoyage AI multimodal embedding API |
Related models
- voyage-multimodal-3.5
Voyage AI · Voyage Multimodal
Shared embeddings for text, images and video frames.
Launch date API - voyage-context-3
Voyage AI · Voyage Context
Document-aware chunk embeddings trained for chunk and document relevance.
Launch date API