What changed for search applications
Amazon S3 Vectors now supports hybrid search, combining vector similarity and full-text matching in one query and returning a single ranked result list. AWS announced the capability on November 16, 2026.
The change is relevant to retrieval-augmented generation, agent memory and search applications that need both semantic meaning and exact terms. A developer can use text matching alongside an embedding instead of relying only on similarity between vectors.
How the existing APIs handle text
Developers include searchable text as a metadata field when writing data with PutVectors. S3 Vectors automatically indexes that field for full-text search using BM25 relevance scoring. Its default analyzer is language-agnostic, with language-specific analyzers also available.
A QueryVectors request can supply both query text and a query vector. S3 Vectors runs the two searches, combines them with Reciprocal Rank Fusion and returns the requested top results. Developers can also query full text and vectors separately, search multiple text fields, or enable full-text search on existing data in place.
Availability and next steps
AWS says the feature is available in all commercial AWS Regions where S3 Vectors is offered, and in AWS China Regions. This is a service-specific availability statement, rather than confirmation that every AWS Region supports it.
Developers can get started through the AWS CLI, SDKs or S3 console and follow the documentation and pricing links in the announcement. The release does not state a new fixed unit price. AWS's broader promotional cost and latency claims are not independent measurements of a particular application's results.



