Turbopuffer — Serverless Vector DB for AI Search

What is Turbopuffer?

Turbopuffer is a serverless vector database with sub-millisecond query latency, automatic scaling, and pay-per-query pricing. No clusters or replicas to manage.

In one sentence: Serverless vector database — sub-millisecond queries, auto-scales to zero, pay-per-query with no minimums. a16z-backed.

For: AI teams building RAG or semantic search.

Core Features

1. Zero Ops

No clusters or replicas — use the API key directly.

2. Attribute Filtering

Supports conditional filters ($eq/$lt/$gt/$in).

3. Hybrid Search

Vector similarity + BM25 text search.

FAQ

Q: Pricing? A: Pay per query and storage — no usage, no cost.

Q: Can I migrate from Pinecone? A: Yes — export vectors then upsert to Turbopuffer.