pgvector vs Pinecone vs Qdrant: When Is a Dedicated Vector Database Actually Worth It?

Short summary

A practitioner's guide to choosing between pgvector, Pinecone, and Qdrant based on real shipping experience. pgvector wins for low-to-mid millions of vectors when data already lives in Postgres; dedicated databases become worth it when recall, concurrency, or operational isolation starts to hurt. The author argues against switching based on hype alone and provides concrete SQL examples and trigger criteria for each transition.

• •pgvector is the right first choice when data is already in Postgres and corpus is in the low millions
• •Switch to Pinecone or Qdrant when recall, scale, or operational isolation becomes painful
• •Pinecone is managed and closed-source; Qdrant is open-source and self-hostable, each with distinct tradeoffs

Generated with AI, which can make mistakes.