v.25.8New Feature

Vector similarity index now supports binary quantization

The vector similarity index now supports binary quantization. Binary quantization significantly reduces the memory consumption and speeds up the process of building a vector index (due to faster distance calculation). Also, the existing setting vector_search_postfilter_multiplier was made obsolete and replaced by a more general setting : vector_search_index_fetch_multiplier. #85024 (Shankar Iyer).
The vector similarity index in ClickHouse now supports binary quantization, enabling more efficient memory usage and faster index building.

Why it matters

Binary quantization reduces the memory consumption of vector similarity indexes and accelerates the building process by enabling faster distance calculations. This improvement optimizes resource utilization and query performance when working with vector data.

How to use it

Users can take advantage of binary quantization by enabling the updated vector similarity index settings. The previous setting vector_search_postfilter_multiplier is now obsolete and replaced by vector_search_index_fetch_multiplier, which can be configured to control index fetching behavior.