At a glance
| Product type | Distributed search and analytics engine with vector search, plus managed services |
| Vector index types | HNSW with optional int8, int4 or BBQ quantization; flat variants; bbq_disk (Enterprise) |
| Hybrid search | RRF and linear retrievers; ES|QL; lexical plus semantic_text in one query |
| Metadata filtering | Filters inside approximate and exact kNN queries |
| Embedding integration | semantic_text with Elastic Inference Service, ELSER or third-party inference endpoints |
| Scaling and multitenancy | Sharded indices across nodes; Serverless projects scale by usage |
| Deployment (managed or self-hosted) | Self-managed, Elastic Cloud Hosted, Elasticsearch Serverless |
| Licensing and pricing model | AGPLv3, SSPL or ELv2 for free features; subscriptions; hosted from 99 USD per month; Serverless per VCU-hour and GB |
Elasticsearch suits teams that want vector search inside a mature full-text engine, with BM25, filters, aggregations and hybrid ranking in one query. Self-managed use is free for core features, but some vector options such as bbq_disk need an Enterprise subscription, so check the subscription matrix before choosing a deployment.
Overview
Elasticsearch is a distributed search engine built on Apache Lucene that stores JSON documents in indices split into shards. For vector search it offers the dense_vector field (up to 4,096 dimensions), approximate kNN on HNSW-based index types with optional scalar or BBQ quantization, exact brute-force kNN, and semantic_text fields that generate embeddings through an inference endpoint. The latest version listed in the release notes on 7 November 2026 was 9.5.6.
Elastic sells Elasticsearch three ways: self-managed with free and paid subscription features, Elastic Cloud Hosted (resource-based pricing on AWS, Google Cloud and Azure), and Elasticsearch Serverless (usage-based pricing). Elastic moved the code from Apache 2.0 to SSPL and the Elastic License with 7.11 in 2021 (OpenSearch is the Apache 2.0 fork that followed) and added AGPLv3 as a licence option in 2024.
Key features
- dense_vector fields with float, bfloat16, byte and bit elements, up to 4,096 dimensions, and l2_norm, dot_product, cosine or max_inner_product similarity
- Index types hnsw, int8_hnsw, int4_hnsw, bbq_hnsw, flat variants and bbq_disk (DiskBBQ, Enterprise subscription)
- Approximate kNN with filters, plus exact brute-force kNN that scores every matching document
- semantic_text fields that create embeddings automatically through the Elastic Inference Service or a configured inference endpoint
- Hybrid search with the RRF retriever, linear retriever and ES|QL combining lexical and semantic matches
- ELSER (Elastic Learned Sparse Encoder), packaged for machine learning nodes, for learned sparse retrieval
- Full-text BM25 search, aggregations and Kibana for exploring and monitoring data
Integrations and ecosystem
Pros and cons
Pros
- Mature full-text search, filtering and aggregations next to vector search
- Several quantization options to reduce vector memory use
- Built-in embedding generation through semantic_text and inference endpoints
- Managed options on AWS, Google Cloud and Azure with published starting prices
Cons
- Some vector features, including bbq_disk, require an Enterprise subscription
- The subscription feature matrix did not render as text, so the tier for RRF and other retrievers could not be confirmed
- Running a self-managed cluster means sizing shards, memory and nodes yourself
- Licence history (Apache 2.0, then SSPL/ELv2, then AGPLv3 added) needs legal review for redistribution
Pricing
Self-managed Elasticsearch has free features plus paid subscriptions (Platinum, Enterprise) licensed by node count and RAM, by quote. Elastic Cloud Hosted is resource-based: the pricing page showed Standard from 99 USD, Gold 114 USD, Platinum 131 USD and Enterprise 184 USD per month for a reference configuration of 120 GB storage across 2 zones. Elasticsearch Serverless is usage-based, from 0.14 USD per ingest VCU-hour, 0.09 USD per search VCU-hour, 0.07 USD per machine learning VCU-hour and 0.047 USD per GB retained per month. A free trial is offered; the hybrid search docs mention 14 days for Serverless.
Pricing checked on the vendor's official page on 7 November 2026. Prices change; confirm before buying.
Who it is best for
Teams that already use Elasticsearch for search or logs, or that need relevance tuning, filters and hybrid lexical and vector ranking in one engine.
Typical use cases
- Hybrid product and site search combining keywords and embeddings
- Retrieval for RAG assistants over large document collections
- Semantic search across support tickets, knowledge bases and logs
- Image or text similarity search with filters on metadata
Who should look elsewhere
- Small applications that only need vector similarity and would be better served by a database extension such as pgvector or a lightweight vector database.
Verdict
Elasticsearch is a strong option when vector search must sit alongside full-text relevance, filters and analytics, and its managed services have clear starting prices. Budget for subscription tiers if you need features such as bbq_disk, and compare OpenSearch if an Apache 2.0 licence matters to you.
Frequently asked questions
Is Elasticsearch open source again?
Elastic's licensing FAQ says the free portions of Elasticsearch are available under AGPLv3, SSPL 1.0 or the Elastic License 2.0 at your option. AGPLv3 was added in 2024 and is OSI-approved; paid features remain under Elastic's commercial terms.
Does Elasticsearch support hybrid search?
Yes. Elastic documents hybrid search with the RRF retriever combining a lexical match and a semantic_text match, a linear retriever, and ES|QL queries that mix keyword and semantic matching.
How much does Elastic Cloud cost?
On 7 November 2026 Elastic Cloud Hosted started at 99 USD per month (Standard) up to 184 USD (Enterprise) for a 120 GB, two-zone reference setup. Serverless is usage-based from 0.09 USD per search VCU-hour and 0.047 USD per GB per month.
What is the maximum vector dimension in Elasticsearch?
The dense_vector field accepts up to 4,096 dimensions, according to the mapping reference.
Sources
- Elastic licensing FAQ
- Elastic Cloud Hosted pricing
- Elasticsearch Serverless pricing
- dense_vector field reference
- kNN search guide
- Hybrid search with semantic_text
- Elasticsearch release notes
Checked November 2026.
Listed in: Vector Databases & AI Search.
How we research reviews and what star ratings mean: our editorial method.