voyage-3.5 and voyage-3.5-lite: improved quality for a new retrieval frontier
a year ago
- #AI
- #Embedding Models
- #Machine Learning
- Introduction of voyage-3.5 and voyage-3.5-lite, the latest embedding models with improved retrieval quality.
- Both models support embeddings in 2048, 1024, 512, and 256 dimensions with multiple quantization options.
- voyage-3.5 and voyage-3.5-lite outperform OpenAI-v3-large by 8.26% and 6.34% respectively, with lower costs.
- Matryoshka learning and quantization-aware training enable reduced vector database costs by up to 83%.
- Evaluation across 100 datasets in eight domains shows superior performance in retrieval quality.
- Binary rescoring with full-precision embeddings improves retrieval quality by up to 6.89%.
- First 200 million tokens are free; available now with documentation and community support.