Qdrant

What it's good for

• 1

  RAG over a corpus that has to stay on the client's own infrastructure

• 2

  Semantic search inside a product where data sovereignty is a hard requirement

• 3

  Large-scale vector workloads (50M+ vectors) where managed pricing gets expensive

• 4

  Self-hosted agent memory backed by retrievable embeddings

• 5

  Hybrid dense-plus-sparse search when keyword and semantic matching both matter

Pricing

• Cloud Free

  1GB RAM cluster, no card required

  Free

• Cloud Paid

  Billed on infrastructure, ~$0.078/GB-hour (about $57/mo per GB RAM)

  Usage-based

• Hybrid Cloud

  Your compute, Qdrant-managed control plane

  Custom

• Private Cloud

  Fully self-managed, air-gap friendly

  Custom

• Self-hosted

  Open source, Apache-2.0

  Free

How to use it

Start on the free cloud tier to prototype, or run the Docker image locally to keep everything in-house. Create a collection with the vector dimension matching your embedding model, then upsert points with payloads (metadata you can filter on). Query with a vector plus a payload filter to combine semantic and structured search. When you outgrow the free tier, decide between paid cloud (billed on infrastructure), Hybrid Cloud (your compute, their control plane), or Private Cloud for fully air-gapped deployments. Pair it with an eval layer like Langfuse before you ship.

Pros & cons

Frequently asked questions

• Is Qdrant free?

  Yes, in two ways. The open-source engine is free to self-host under an Apache-2.0 license, and the managed cloud has a free tier with a 1GB-RAM cluster that needs no credit card. Paid cloud is billed on the underlying infrastructure at roughly $0.078 per GB-hour, which lands around $57/mo per GB of RAM.

• Qdrant vs Pinecone: which should I pick?

  Pinecone is fully managed and gets you to production RAG fastest with the least ops. Qdrant is open source, so you can self-host it for data sovereignty or air-gapped deployments, and it tends to come in cheaper past tens of millions of vectors. If a client demands they own the stack or keep data on their own hardware, Qdrant wins. If they want to rent a hands-off service, Pinecone wins.

• Is Qdrant open source?

  Yes. The core vector database is open source under an Apache-2.0 license, so you can download it, read the code, and run it on your own hardware with no vendor lock-in. The managed cloud is a paid convenience layer on top of that same engine.

• What is Qdrant best for?

  Production vector search where you need to own the stack: RAG and semantic search that must stay on your own infrastructure, and large-scale workloads where fully-managed pricing gets steep. The Rust engine handles real throughput, and the self-host option keeps sensitive data in-house.

• Does Qdrant get cheaper than managed vector databases?

  At scale, usually. Around 50 million vectors and up, self-hosting Qdrant or running its infrastructure-billed cloud tends to cost less than per-request managed pricing. Below that, the convenience of a fully-managed service can be worth the premium. Model your own vector count and query volume before deciding.

More

• Qdrant docs
• Qdrant Cloud pricing

Alternatives to Qdrant

Other tools we'd consider for the same job.

• Pinecone

  Fully-managed vector database for production RAG. Serverless tier, low operational overhead — the default for client deployments

Other Data & ML Ops tools

• Langfuse

  Open-source LLM engineering platform: tracing, evals, and prompt management. The own-your-traces answer to LangSmith

• Ollama

  Free, open-source runtime for local LLMs. Run Llama, DeepSeek, Mistral and more on your own machine with one command

• Braintrust

  Commercial eval and observability platform for LLM apps: evals, logging, and a prompt playground with unlimited users

• Pinecone

  Fully-managed vector database for production RAG. Serverless tier, low operational overhead — the default for client deployments