Qdrant · 2× · naposledy 13. 8. 2026
69
Momentum
Qdrant is an Open Source vector database and similarity search engine written in Rust for embedding-based applications such as semantic search, RAG, and recommendation systems. It supports dense, sparse, and multi-vector embeddings, hybrid search, and filtering via payload metadata. Qdrant is available as a self-hosted Open Source version (Apache 2.0) as well as managed Qdrant Cloud (Managed, Hybrid, Private Cloud). The current version 1.19 introduces prefix matching on keyword indices, a new TurboQuant 4-bit data format, and unified storage tiers.
Vývoj momenta
15.05.13.08.
Vlastnosti
| Throughput/Latency | Up to 4x higher RPS vs. competitors per own benchmarks; quantization (scalar/product/binary) reduces memory and boosts search performance up to 40x |
| License | Apache License 2.0 (open source) |
| Platform | Docker container, Linux/cloud, Web UI, clients for Python, JavaScript/TypeScript, Rust, Go, Java, .NET |
| Price | Free self-host (Apache 2.0); Cloud free tier with 1GB RAM/4GB disk; Standard tier usage-based (vCPU, RAM, disk, inference tokens), billed hourly |
| Protocol Compatibility | REST API (OpenAPI 3.0) and gRPC interface for production-grade, high-performance access |
| Release Date | Version 1.19.0 released August 4, 2026 |
| Supported Models/Providers | Cloud Inference (externally hosted): OpenAI, Cohere, Jina AI, OpenRouter; also integrations with Bedrock, Mistral, Aleph Alpha, Gemini; curated own models like all-MiniLM-L6-v2, splade-pp-en-v1, bm25 |