Binary Quantization & Rescoring: 96% Less Memory, Faster Search

MongoDB has released new capabilities for binary quantized vector ingestion, automatic scalar quantization, and automatic binary quantization and rescoring in public preview. These enhancements empower developers to scale semantic search and generative AI applications more cost-effectively by reducing memory usage and improving scalability. The new "quantization" index definition parameters allow developers to choose between full-fidelity vectors or quantized vector embeddings with a balance of storage efficiency and search accuracy. Automatic rescoring is incorporated when using binary quantization, ensuring highly accurate final search results despite initial vector compression. This enables the efficient processing of massive knowledge bases for analysis and insight-oriented use cases, such as content summarization and sentiment analysis, as well as retrieval-augmented generation applications and A/B testing of different embedding models.