Source: Weaviate docs: Hybrid search

Weaviate documentation for hybrid BM25 + vector queries: alpha weighting, fusion methods and thresholds.

Contents
  • What: vendor documentation
  • Author: Weaviate
  • Date: read 2026-11-14
  • Original: Hybrid search

Summary

Reference for running a keyword (BM25) search and a vector search in one query and fusing the results.

Key claims

  • alpha weights the two sides: 1 is pure vector, 0 is pure keyword.
  • Two fusion methods: Ranked Fusion (by rank) and Relative Score Fusion (by normalized score); Relative Score Fusion is the default since v1.24.
  • maxVectorDistance is the only threshold and applies only to the vector side.
  • queryProperties restricts and boosts (e.g. question^2) the fields the keyword side searches.

Feeds

Hybrid search

Suggest changes →