turbopuffer is a serverless vector and full-text search database built on object storage, priced from $16 per month.
turbopuffer is a managed search database hosted on object storage: writes are made durable in a write-ahead log, warm queries are served from an NVMe cache, and vector approximate-nearest-neighbour ranking and BM25 full-text ranking run inside the same namespace. Hybrid queries fan out a vector and a full-text sub-query and fuse the results server-side with reciprocal rank fusion.
Self-serve plans start at $16 per month for Launch and $256 per month for Scale, billed on stored bytes, writes, and queries; the Enterprise tier starts at $4,096 per month and adds single-tenancy and a 99.95% uptime SLA. A namespace scales to 128 billion documents at 256 TB, up to 256 namespaces can be pinned warm, and a large namespace is sharded across storage nodes automatically.
