Which vector database fits us, Pinecone, Qdrant or Weaviate?

For SMB RAG under a million chunks, search-quality differences are small. The decision is about hosting model, ops cost and stack fit. pgvector is often the boring but correct option if you already run Postgres.

Try this first

1. 1pgvector on Postgres: pick this if you already run Postgres (Neon, Supabase, RDS). No extra vendor, one backup strategy, fine up to a few hundred thousand chunks.
2. 2Qdrant: open-source, self-host or cloud, strong filters, good docs. Pick this when you want control or for on-prem without Postgres dependency.
3. 3Pinecone: SaaS only, zero ops, scales well. Pick this if you do not want to spend ops time and the cost fits. Not for on-prem or strict non-US data location.
4. 4Weaviate: open-source with SaaS option, strong on hybrid search (keyword plus vector). Pick this if you combine heavy filtering with text search.
5. 5Do not only count licenses. Count hosting hours, storage cost and developer maintenance time. At 10,000 chunks everything is cheap, at a million it starts mattering.

None of the above fits?

Describe your situation below. We pass your input plus the steps you already saw to our AI and return tailored next-step advice. If it's too risky to DIY, we'll say so.