pgvector vs turbopuffer
Comparing two vector database platforms on pricing, features, free tier, and trade-offs.
Quick summary
pgvector — Vector similarity search for Postgres. pgvector is an open-source Postgres extension that adds vector similarity search to any Postgres database. Runs anywhere Postgres runs (Supabase, Neon, RDS).
turbopuffer — Serverless vector search on object storage. turbopuffer is a serverless vector database built on S3, offering very cheap storage pricing and pay-per-query model — designed for RAG at scale without fixed pod costs.
Feature comparison
| Feature | pgvector | turbopuffer |
|---|---|---|
| Pricing model | Free | Paid |
| Starting price | Free (just Postgres) | Usage-based |
| Free tier | Yes | No |
| Open source | Yes | No |
| Type | Postgres extension | Serverless |
| Free Tier | Unlimited | None |
| Serverless | No | Yes |
| Self-hosted | Yes | No |
| Multi-tenant | Yes | Yes |
| Hybrid Search | Yes | Yes |
| Max Dimensions | 16000 | 10000 |
| Metadata Filtering | Yes | Yes |
pgvector
Vector similarity search for Postgres
Pros
- Use your existing Postgres
- No new infrastructure
- Transactional guarantees with vectors
- Free — pay only for Postgres
Cons
- Slower than purpose-built vector DBs at scale
- Index build times grow with data
- Not ideal for 100M+ vectors
turbopuffer
Serverless vector search on object storage
Pros
- Storage on S3 — extremely cheap
- Pay per query, no pod hours
- Good for cold / infrequently-queried data
- Simple API
Cons
- Higher query latency than Pinecone/Qdrant
- No free tier
- Closed source
Which should you choose?
Choose pgvector if you value open source and want the option to self-host, and a free tier is important for your stage. Choose turbopuffer if you need production-grade features and are ready to pay.