Vector Databases
Vector databases index high-dimensional embeddings so you can retrieve semantically similar text, images, or code in milliseconds — the storage layer behind most RAG and semantic-search systems. The main trade-offs are managed vs self-hosted, whether you need hybrid (keyword + vector) search, and how pricing scales with vectors stored and queries served.
| Chroma Developer-friendly open-source embedding database. | Open source (Apache-2.0) | Managed, Self-hosted | Open source (free) | No | Cloud | Yes |
| LanceDB Embedded, serverless vector database on the Lance columnar format. | Open source (Apache-2.0) | Self-hosted, Managed | Open source (free) | Yes | Yes | Yes |
| Milvus / Zilliz Scalable open-source vector database for billion-scale search. | Open source (Apache-2.0) | Managed, Self-hosted | Open source (free) | Yes | Cloud | No |
| MongoDB Atlas Vector Search Vector search built into MongoDB Atlas documents. | Source available (MongoDB core is SSPL; Atlas Vector Search is a managed Atlas feature) | Managed | Free tier + paid | Yes | Atlas | No |
| pgvector Vector similarity search as a Postgres extension. | Open source (PostgreSQL License) | Self-hosted, Managed | Open source (free) | Yes | Via Postgres host | No |
| Pinecone Fully managed, serverless vector database. | Proprietary | Managed | Free tier + paid | Yes | Yes | Yes |
| Qdrant Fast open-source vector search engine written in Rust. | Open source (Apache-2.0) | Managed, Self-hosted | Open source (free) | Yes | Cloud | No |
| Redis (Vector) Vector search on the in-memory data platform. | Source available (RSALv2/SSPLv1; AGPLv3 available from Redis 8) | Self-hosted, Managed | Free tier + paid | Yes | Cloud | No |
| turbopuffer Serverless vector search built on object storage for low cost. | Proprietary | Managed | Usage-based | Yes | Yes | No |
| Vespa Open-source engine for large-scale search, ranking, and retrieval. | Open source (Apache-2.0) | Self-hosted, Managed | Open source (free) | Yes | Cloud | Yes |
| Weaviate Open-source vector database with built-in hybrid search and modules. | Open source (BSD-3-Clause) | Managed, Self-hosted | Open source (free) | Yes | Cloud | Yes |
Compiled from public docs and vendor sites; verify current pricing with the vendor. No vendor pays for placement. Each tool links to a fuller profile.
How to choose
Pick managed (Pinecone, Zilliz, Turbopuffer) if you want zero ops and predictable scaling; pick open-source (Qdrant, Weaviate, Milvus, Chroma) if you want to self-host or avoid lock-in. If you already run Postgres, pgvector avoids a new system entirely. Check hybrid search and metadata filtering if your retrieval needs keywords plus vectors.
FAQ
Do I need a dedicated vector database?
Not always. If you already run Postgres, pgvector handles millions of vectors well. Dedicated engines pay off at larger scale, for lower query latency, or when you need advanced filtering and hybrid search.
What is hybrid search?
Hybrid search combines keyword (BM25/sparse) and vector (dense) retrieval, then fuses the results. It improves recall for queries with exact terms like product codes or names that pure semantic search can miss.
How is pricing usually structured?
Managed services typically bill on vectors stored, dimensions, and queries or on provisioned pods/compute. Open-source engines are free to run; you pay for the infrastructure and your own operations time.