Qdrant 2.0: Open-Source Vector DB Innovations
Qdrant launches version 2.0 with Discovery API, sparse vectors, and doubled performance. Overview of new features for your RAG applications.
Qdrant 2.0: A Major Version
Qdrant, the open-source vector database, reaches a milestone with version 2.0. This release brings long-awaited features and significant performance improvements.
"Qdrant 2.0 represents 18 months of intensive development," declares Andrey Vasnetsov, CEO of Qdrant. "We listened to the community and delivered the most requested features."
New Features
Discovery API
The Discovery API enables interactive data exploration:
- Navigation by similarity
- Cluster exploration
- Neighborhood visualization
This feature opens new use cases for semantic search.
Native Sparse Vectors
Qdrant now supports sparse vectors natively:
| Type | Use Case | Performance |
|---|---|---|
| Dense | Semantic questions | Excellent |
| Sparse | Technical terms | Very good |
| Hybrid | General use | Optimal |
The dense + sparse combination implements hybrid search directly in the database.
Query Groups
Result grouping by attribute:
- Aggregation by category
- Intelligent deduplication
- Optimized pagination
Performance
Benchmarks
Qdrant 2.0 doubles performance on several metrics:
| Metric | v1.x | v2.0 | Improvement |
|---|---|---|---|
| QPS | 450 | 920 | +104% |
| P50 Latency | 15ms | 8ms | -47% |
| P99 Latency | 65ms | 32ms | -51% |
Optimizations
Gains come from:
- New optimized HNSW index
- Improved memory compression
- Increased parallelism
Deployment
Docker
DEVELOPERbashdocker run -p 6333:6333 qdrant/qdrant:v2.0.0
Kubernetes
DEVELOPERbashhelm install qdrant qdrant/qdrant --version 2.0.0
For a complete guide, check out our advanced Qdrant tutorial.
Migration from 1.x
Migration is transparent:
- Collection backup
- Binary upgrade
- Feature verification
APIs remain compatible, no code changes required.
Comparison with Competition
| Feature | Qdrant 2.0 | Pinecone | Weaviate |
|---|---|---|---|
| Open source | Yes | No | Yes |
| Sparse vectors | Native | Via API | Plugin |
| Discovery API | Yes | No | No |
| Auto-scaling | Cloud | Yes | Yes |
Qdrant Cloud Pricing
| Plan | Price | Included |
|---|---|---|
| Free | Free | 1GB, 1 collection |
| Pro | $25/month | 10GB, unlimited |
| Enterprise | Custom | SLA, support |
Conclusion
Qdrant 2.0 confirms its position as open-source leader. For teams prioritizing control and flexibility, it's an excellent choice.
Explore our comparison of vector databases to choose the best option.
Need to integrate Qdrant? Ailog uses Qdrant as a backend to offer performance and data sovereignty. Deploy your RAG in minutes.
Tags
Related Posts
Vector Databases 2026: Trends and New Players
Complete overview of the vector database market in 2026. New entrants, major evolutions, and comparison of solutions for your RAG applications.
Pinecone Serverless: Evolutions and Pricing
Pinecone announces major updates to its Serverless offering: new features, price reductions, and improved performance.
Llama 4: Open Source Catches Up with Proprietary Models
Meta unveils Llama 4 with RAG performance rivaling GPT-5 and Claude 4. Open source crosses a decisive threshold for enterprise applications.