Qdrant 1.20.1: 50% fewer cold-storage reads, ~2x faster quantization, agent-ready setup

qdrant_engine · x · 2026-11-16

Vector database Qdrant ships 1.20.1 with notable performance gains: 50% fewer reads during HNSW graph search on cold storage, 2x faster Turbo8/Turbo16 quantization, faster HNSW graph construction via PiPNN, plus faster filtering, batch searches, shard transfers and record streaming. It also improves agent integration: an llms.txt endpoint for one-line coding-agent setup, with Agent Skills, Cloud CLI and MCP support.

Original post →

More from Infra

Infra channel →