pgvector vs Qdrant vs LanceDB

Three vector stores

All three answer nearest-neighbor queries. The split is the rest of the application: SQL joins, payload filters on a dedicated engine, or blobs in a Lance table.

pgvector

pgvector is open-source vector similarity search for Postgres: vector columns, exact search, and HNSW or IVFFlat indexes, with SQL joins and filters on the rest of the row. Any host that runs the extension (self-managed Postgres, Supabase, Neon, and others) can serve it. Media usually lives in object storage you add. Repo: https://github.com/pgvector/pgvector

Qdrant

Qdrant is an open-source vector search engine written in Rust. Collections hold points: vectors plus payload. Payload indexes extend the HNSW graph so filters apply during search. It supports dense and sparse vectors, BM25 on a text field, and deployments from OSS to Qdrant Cloud, Hybrid Cloud, and Private Cloud. Docs: https://qdrant.tech/documentation/overview/

LanceDB

LanceDB stores images, video, audio, text, embeddings, and features in one Lance table and serves vector search, full-text search, hybrid search, reranking, filtering, and SQL. OSS is an embedded library. Enterprise is a distributed lakehouse. Docs: https://docs.lancedb.com/

How to choose

Start from the system of record you already run.

• Pick pgvector when Postgres already holds the application row and SQL is how you join and filter.
• Pick Qdrant when a dedicated ANN service with rich payload filters is the serving contract.
• Pick LanceDB when multimodal blobs and vectors should share one Lance table, embedded or as a lakehouse.

Last reviewed

Facts on this page were checked against vendor documentation on 2026-10-28. Product surfaces move; treat version names and APIs as directional and follow the official links in Related.

pgvector vs Qdrant vs LanceDB (last reviewed: Oct 2026)

Where Pixeltable fits

Use Pixeltable when the index should be declared on the table that holds the passage or frame, so insert and delete keep it current, and typed media can sit on the same schema. Pixeltable is not a drop-in Qdrant cluster and not a Lance lakehouse. Keep pgvector when Postgres is already the system of record and you will embed in a worker you write. Keep Qdrant when that engine is the serving contract. Keep LanceDB when Lance layout is the product; export with pxt.io.export_lancedb when you want that engine.

import pixeltable as pxt
from pixeltable.functions.huggingface import sentence_transformer

TableModel = pxt.model_base()
embed = sentence_transformer.using(
    model_id='sentence-transformers/all-MiniLM-L6-v2'
)

class Passages(TableModel, name='passages'):
    text: pxt.String
    source: pxt.String
    __indexes__ = [
        pxt.EmbeddingIndex(text, string_embed=embed),
    ]

@pxt.query
def search_passages(query: str, limit: int = 5):
    sim = Passages.text.similarity(string=query)
    return (
        Passages.order_by(sim, asc=False)
        .limit(limit)
        .select(Passages.text, Passages.source, sim)
    )

Questions