Qdrant is written in Rust specifically for performance, and its core strength is filtered search: running vector similarity search alongside metadata filters, like date ranges or category tags, without the usual trade-off where adding filters slows a query down significantly. It also supports hybrid search combining dense vector embeddings with sparse retrieval methods, similar in spirit to Weaviate's approach but with different internals.
Self-hosting Qdrant has no feature gating between the free and paid tiers, meaning anyone running their own cluster gets the same capabilities as a paying Qdrant Cloud customer, just without the managed infrastructure. Qdrant Cloud itself bills hourly based on resource usage rather than a flat subscription, which can work out cheaper for spiky workloads and more expensive for steady, predictable ones.
It has a free tier to start and paid tiers beginning around $25 a month. Among the open-source vector databases, Qdrant has built a reputation specifically around raw query speed, which matters most for applications running vector search at high volume or with tight latency requirements.