voyage-3
Voyage 3 is the general third-generation text embedding model. It expanded the input limit from the earlier voyage-2 generation.
Model design
The release pairs a 32,000-token input limit with 1,024-dimensional output. It evaluates general and multilingual retrieval. The separate Lite release reduces vector size as well as targeting lower inference cost.voyage-3 and voyage-3-lite
Practical use
Practical guidance. A longer input limit can reduce forced truncation, but an entire document vector can obscure a short relevant passage. Compare full-document and passage retrieval before deciding how to divide the source text.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API |
| Access | Earlier Voyage API modelVoyage AI text embedding catalog and earlier models |
Related models
- voyage-3.5
Voyage AI · Voyage 3.5
Updated general embeddings distilled from larger retrieval models.
Launch date API - voyage-3-lite
Voyage AI · Voyage 3
Smaller third-generation encoder with 512-dimensional output.
Launch date API Earlier general text encoder with a 4,000-token limit.
Launch date not confirmedAPI