Amazon S3 Vectors increases filter data to 40 KB for each vector
The new limit is 20 times the limit of 2 KB before, and it has no cost.
Amazon says that Amazon S3 Vectors increased its limit of filter metadata to 40 KB for each vector, from 2 KB. S3 Vectors stores vectors, which are sequences of numbers that help AI find items with almost the same meaning. Metadata is more data on a vector, such as category, permissions and tags. A query can use more of these fields to narrow its results.
What changes
New indexes in new vector buckets have the larger limit by default. The procedure to write vectors and to make queries with a filter does not change. An index that you made before must change to enhanced index mode with the UpdateIndexMode API (an interface for software).
The proof
The source gives no results for speed or precision. It says that the larger limit has no cost in all AWS Regions where S3 Vectors is available.
What is not known
The source does not say if enhanced index mode changes other parts. It gives no price for the vector storage.
Insight: Vector search is moving into storage
Amazon raised the S3 Vectors limit from 2 KB to 40 KB at no cost. In November 2026, Onehouse said that its vector search index is in the same bucket as open lake tables. This needs no data copy. This reads as a move to make storage the area for vector search, and a different vector database is not easy to sell. Monitor which tools keep data in open formats.
Sources
Posted