VectorChord

VectorChord is a PostgreSQL extension for scalable, high-performance vector similarity search — the successor to the popular pgvecto.rs project. Built around the RaBitQ quantization algorithm, it is designed for the billion-scale era: its benchmarks host 100M vectors on a single modest instance and scale past a billion, while indexing many times faster than pgvector. It is an open-source project on GitHub with around 1,800 stars, under very active development in 2026.

Overview

RAG and semantic-search workloads usually meet Postgres through pgvector, whose memory-hungry indexes strain at large scale. VectorChord attacks the cost side: RaBitQ compresses vectors to 4-bit or 8-bit representations with minimal recall loss, indexes can be built on a powerful machine and queried on a smaller one, and data lives on disk efficiently instead of demanding huge RAM. It speaks the familiar Postgres extension model — create an index, then query with standard distance operators — so existing Postgres applications gain serious vector capacity without adopting a separate vector database.

Key Features

• PostgreSQL-native vector search, successor to pgvecto.rs
• RaBitQ 4-bit/8-bit quantization for disk- and memory-efficient indexes
• Benchmarks far beyond pgvector in index speed and insert throughput
• Scales from millions to over a billion vectors
• Build indexes on big machines, serve queries on small ones
• Standard SQL interface with familiar distance operators

💳 Pricing

Free. VectorChord is open source and self-hosted as a Postgres extension; managed offerings exist through cloud partners, but the extension itself costs nothing.

Who Is It For

Developers building RAG, semantic search and recommendation features on PostgreSQL whose vector counts have outgrown pgvector’s practical limits.