voyage-multilingual-2
Voyage Multilingual 2 targets retrieval across languages. It handles multilingual documents and queries in a shared embedding system.
Model design
The release extends the second-generation catalog beyond its earlier general and domain-specific models. It supports 32,000-token inputs and 1,024-dimensional output. English retrieval remains part of its stated evaluation focus.voyage-multilingual-2
Practical use
Practical guidance. Build queries in the languages that users actually use, including mixed-language queries. Check cross-language retrieval separately from same-language retrieval. A combined score can hide cases where the right source exists only in another language.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API |
| Context window | 32,000 tokens Per input text. voyage-multilingual-2 |
| Access | Earlier Voyage API modelVoyage AI text embedding catalog and earlier models |
| Output dimensions | 1,024voyage-multilingual-2 |
Related models
General text embeddings compatible with the other Voyage 4 models.
Launch date APIThird-generation general text embeddings with a 32,000-token input limit.
Launch date API