Pinecone Announces General Availability of Full-Text Search in Its Vector Database

On Sep 9, 2026 Pinecone announced that its full‑text search capability is generally available in the Pinecone Database. After months of private preview validation with user partners, the feature adds BM25 keyword ranking, Lucene‑style query syntax, fuzzy matching, and tokenization and stemming in 18 languages. The new functionality is integrated into the same index schema used for dense and sparse vectors, allowing a single index to store text fields alongside vector embeddings. Users can combine exact‑match identifier queries (e.g., SKUs, error codes) with semantic search in one query, and apply text‑match filters to restrict results. The service retains Pinecone’s usage‑based capacity model, metered in read and write units, with optional dedicated read nodes and a bring‑your‑own‑cloud deployment option. Pinecone’s blog post notes that the addition eliminates the need for separate lexical search engines, reducing operational overhead for teams that previously maintained node‑based search clusters. The feature is now accessible through the Documents API and can be enabled by creating an index with a document schema that includes text, dense, and sparse fields.