# Does Zilliz Cloud Blackwell support hybrid vector-keyword search queries?

Source: https://zilliz.com/ai-faq/does-zilliz-cloud-blackwell-support-hybrid-vectorkeyword-search-queries
Author: zilliz

Yes, Zilliz Cloud enables hybrid search combining vector similarity and keyword matching at Blackwell GPU speed, improving retrieval recall and ranking quality.

## Dual-Path Retrieval

Zilliz Cloud executes vector and keyword search in parallel on Blackwell. Vector search retrieves semantically-similar documents; keyword search finds exact term matches. Results merge by relevance. Documents missing from vector-only retrieval get second chance via keywords.

## Unified Ranking

Hybrid results rank by combined score: vector similarity normalized + keyword match weight + metadata signals. Users tune weights per use-case. Recommendation systems can weight similarity heavily; compliance searches weight keyword match.

## Sparse-Dense Combination

Zilliz Cloud stores both dense embeddings (vector) and sparse keyword indexes simultaneously. Queries hit both indices at GPU speed. Storage overhead is minimal; performance benefits are substantial.

## Real-World Relevance

For domain-specific search (legal, medical), keyword matching catches terminology that embeddings might miss. Combining both methods retrieves higher-quality results with fewer false negatives.

## Related Resources

- [What is a vector database](https://zilliz.com/learn/what-is-vector-database)
- [Semantic search](https://zilliz.com/glossary/semantic-search)
- [Zilliz Cloud managed vector database](https://zilliz.com/what-is-zilliz-cloud)