voyage-3-large
Voyage 3 Large adds flexible output representations to the third-generation family. It supports Matryoshka dimensions and quantization-aware embeddings.
Model design
The release evaluates retrieval across eight domains and a separate multilingual set spanning 26 languages. Output can use floating-point, integer or binary representations. A reduced representation changes the storage and retrieval trade-off; headline results do not apply equally to every format.voyage-3-large and flexible output representations
Practical use
Practical guidance. Record the selected representation with the index. Compare recall after changing both dimensions and precision, since each can affect ranking. Do not assume vectors from a newer model can be inserted into an older index without re-embedding.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API |
| Context window | 32,000 tokens Per input text. voyage-3-large and flexible output representations |
| Atlas lifecycle | Legacy since 15 January 2026MongoDB Voyage AI model lifecycle: Atlas scope and release history |
| Output dimensions | 1,024 default; 256, 512 or 2,048 optionalMongoDB Voyage AI model specifications |
Related models
- voyage-4-large
Voyage AI · Voyage 4
Mixture-of-experts text embeddings with a shared Voyage 4 vector space.
Launch date API - voyage-3.5
Voyage AI · Voyage 3.5
Updated general embeddings distilled from larger retrieval models.
Launch date API - voyage-3-m-exp
Voyage AI · Voyage 3 experimental
Experimental MTEB-oriented encoder documented through an API model card.
Launch date not confirmedResearch releaseAPI