Voyage 3.5 Embedding Model
Version: 1
Voyage 3.5 is a state-of-the-art text embedding model that delivers superior retrieval quality for general-purpose and multilingual applications. The model outperforms OpenAI-v3-large by 8.26% on average across evaluated domains while offering 2.2x lower costs, making it an optimal choice for organizations seeking high-performance embeddings with cost efficiency.
Voyage 3.5 supports flexible embedding dimensions (256, 512, 1024, and 2048) enabled by Matryoshka learning and offers multiple quantization options including 32-bit floating point, signed and unsigned 8-bit integer, and binary precision. With a 32,000 token context length and multilingual capabilities spanning 26 languages, the model is designed for large-scale retrieval and RAG applications with robust performance across diverse domains including finance, legal, and code.
Model Specifications
LicenseCustom
Last UpdatedNovember 2025
Input TypeText
Output TypeText
ProviderVoyage AI
Languages1 Language