What should I use instead of Elasticsearch for vector search?

If you’re looking to replace Elasticsearch specifically for vector search, the best alternative depends on whether you want:

• pure vector search
• hybrid search (keyword + vector)
• full text search plus vectors
• open-source vs managed

Good alternatives

1. PostgreSQL + pgvector

Best if you want something simple, familiar, and already use Postgres.

• Pros:
  • Easy to adopt
  • Great for small to medium vector workloads
  • Supports hybrid filtering with SQL
  • Good for prototypes and production if scale is moderate
• Cons:
  • Not as fast or scalable as dedicated vector databases at very large scale

2. Qdrant

Best open-source vector database for a lot of teams.

• Pros:
  • Built specifically for vector search
  • Strong filtering
  • Easy API
  • Good performance
  • Can be self-hosted or managed
• Cons:
  • Not a full search engine like Elasticsearch

3. Weaviate

Good if you want vector search plus schema-rich features.

• Pros:
  • Vector + hybrid search
  • Good metadata filtering
  • Flexible and developer-friendly
• Cons:
  • Can feel heavier than Qdrant for simpler use cases

4. Milvus

Best for large-scale vector search.

• Pros:
  • Very scalable
  • Good for high-dimensional embeddings and large datasets
  • Popular in production at scale
• Cons:
  • More operational complexity

5. OpenSearch

Best if you want something closest to Elasticsearch.

• Pros:
  • Familiar if you know Elasticsearch
  • Supports vector search
  • Full-text search + vector search in one system
• Cons:
  • Similar operational model and complexity to Elasticsearch

6. Pinecone

Best managed option if you want to avoid ops.

• Pros:
  • Easy to use
  • Managed, scalable
  • Strong vector-search focus
• Cons:
  • Proprietary/managed service
  • Can be expensive at scale

---

Quick recommendation

• Already using PostgreSQL? → use pgvector
• Want open-source vector DB? → use Qdrant
• Need large-scale production vector search? → use Milvus
• Want Elasticsearch-like hybrid search? → use OpenSearch
• Want fully managed SaaS? → use Pinecone
• Need rich hybrid search with metadata? → consider Weaviate

If you tell me:

1. your dataset size,
2. whether you need keyword + vector hybrid search,
3. whether you want self-hosted or managed,

…I can recommend the best option more precisely.