pgvector vs Qdrant: Postgres extension or open-source vector engine
pgvector adds vector types and HNSW and IVFFlat indexes to an existing Postgres database under the PostgreSQL License, while Qdrant is an Apache-2.0 vector search engine written in Rust with dense, sparse and multi-vector retrieval, BM25 and quantization in one query, so pgvector fits teams that want vectors beside relational data and Qdrant fits teams that need a dedicated, tunable retrieval engine.
Checked
Both are open source and both can be self-hosted, which makes this a different decision from pgvector vs a closed managed service. The question is whether retrieval belongs inside the application's Postgres database or in a separate engine built only for search.
For agents, the deciding factors are usually filtering on metadata, hybrid keyword and vector search, memory cost at the expected corpus size, and whether the team wants one database to operate or two.
What each is
pgvector
A Postgres extension for Postgres 13 and later with vector, halfvec, bit and sparsevec types, HNSW and IVFFlat indexes, and Postgres's transactions, point-in-time recovery and joins. Iterative index scans, added in 0.8.0, address over-filtering.
Qdrant
An open-source vector search engine written in Rust. Queries can combine dense, sparse and multi-vector retrieval, filters and custom scoring. Qdrant 1.20, released on 22 October 2026, added BM25 scoring without sparse vectors, a code tokenizer and TurboQuant 16-bit and 8-bit types. Qdrant raised a $50M Series B on 12 March 2026.
Hosting and pricing model
pgvector
Free under the PostgreSQL License; it runs wherever Postgres runs, and the cost is the database host.
Qdrant
Self-hosted for free, or Qdrant Cloud: a free-forever tier with 0.5 vCPU, 1 GB RAM and 4 GB disk, usage-based Standard billed hourly on vCPU, memory, storage and backups, a Premium tier with a minimum spend, plus Hybrid Cloud and Private Cloud. The pricing page publishes no dollar rates.
Which fits which job
pgvector fits
Applications already on Postgres, retrieval that joins or filters on relational data, and teams that want one database to back up and secure.
Qdrant fits
Retrieval-heavy agents that need hybrid, multi-vector or heavily filtered queries, large corpora where quantization matters, and teams that want search to scale separately from the transactional database.
How Motiontic approaches it
Motiontic's engineers are vendor-neutral. They load a sample of the client's own documents into the candidate stores, run the real questions through each with the same embedding model and reranker, compare recall, latency and cost per query against an eval set, and hand that eval set over so a later migration is a measured decision.
pgvector and Qdrant, as of 30 October 2026
| pgvector | Qdrant | |
|---|---|---|
| What it is | Extension inside Postgres | Standalone vector search engine, written in Rust |
| Licence | PostgreSQL License | Apache 2.0 |
| Index and vector types | HNSW, IVFFlat; vector, halfvec, bit, sparsevec | Dense, sparse and multi-vector in one query |
| Keyword search | Postgres full-text search; BM25 via extensions such as pg_textsearch | BM25 on full-text indexes since 1.20 (22 Oct 2026) |
| Compression | halfvec and bit types | Quantization, including TurboQuant 16-bit and 8-bit types |
| Transactions and joins | Yes, inherited from Postgres | No; a separate store kept in sync |
| Hosting | Any Postgres host | Self-hosted, Qdrant Cloud, Hybrid Cloud, Private Cloud |
| Pricing model | Free; pay for the Postgres host | Free tier; usage-based Standard; no published dollar rates |
| Version and activity | 0.8.7 (1 Oct 2026); about 23.5k GitHub stars | 1.20 (22 Oct 2026); about 35.3k GitHub stars |
Common questions
Is Qdrant open source?
Yes. Qdrant is licensed under Apache 2.0 and can be self-hosted. Qdrant Cloud is the managed service, with a free-forever tier.
Does Qdrant support hybrid search?
Yes. Qdrant combines dense, sparse and multi-vector retrieval with filters, and version 1.20 added BM25 scoring on full-text indexes without needing sparse vectors.
Can pgvector do hybrid search?
Yes, by combining pgvector's similarity search with Postgres full-text search, or with a BM25 extension such as Tiger Data's pg_textsearch, and fusing the two result lists.
How much does Qdrant Cloud cost?
The free tier costs nothing. Standard is usage-based and billed hourly on vCPU, memory, storage and backups, and Qdrant's pricing page points to a calculator rather than publishing dollar rates.
Related
AI agent development
Agents that do real work for users, with the evals, traces and guardrails to keep them in production.
pgvector vs Pinecone
A Postgres extension compared with a managed serverless vector database.
Vector database
Storage and nearest-neighbour search for embeddings, now in many shapes.
Vector quantization
Fewer bits per dimension: smaller, cheaper indexes at a small cost in recall.
HNSW
The layered graph index behind most vector search.
Sources
- pgvector README (index types, vector types, Postgres versions)
- pgvector changelog (0.8.7, 2026-10-01)
- PostgreSQL news, pgvector 0.8.0 (iterative index scans)
- Tiger Data, pg_textsearch 1.0 (2026-03-31)
- Qdrant pricing
- Qdrant 1.20 release (2026-10-22)
- Qdrant, Series B announcement (2026-03-12)
- GitHub, qdrant/qdrant (licence, stars)
Running agents that have to hold up?
Describe the agent, the systems it touches and what it must never get wrong. The first call is with a lab engineer, who will say plainly whether it is ready for production, and whether Motiontic is the right team.