Qdrant released version 1.19 featuring Turbo4 4-bit storage compression (a 9x reduction in footprint), unified memory tiers, and per-tenant IDF calculations for isolated multi-tenant hybrid search.

Key Takeaways

  • ✓Turbo4 stores only 4-bit quantized representations, slashing storage requirements by 9x with high recall;
  • ✓Per-tenant IDF calculates BM25 term rarity exclusively within tenant boundaries to prevent cross-tenant data leakage;
  • ✓Deprecates legacy endpoints in favor of the unified high-throughput `/query` API surface.