At a glance
| Product type | Open source search and analytics engine with a vector plugin |
| Vector index types | HNSW (Faiss, Lucene), IVF (Faiss), flat (Lucene), JVector; NMSLIB deprecated |
| Hybrid search | hybrid query plus normalization or RRF score ranker processors (since 2.11) |
| Metadata filtering | Efficient k-NN filtering or post-filtering |
| Embedding integration | Ingest pipelines with registered models or connectors to externally hosted models |
| Scaling and multitenancy | Sharded indices across nodes; Amazon OpenSearch Service supports up to 1,002 data nodes |
| Deployment (managed or self-hosted) | Self-hosted; Amazon OpenSearch Service domains or Serverless; third-party managed services |
| Licensing and pricing model | Apache 2.0, free; managed services billed by instance hours, OCUs and storage |
OpenSearch suits teams that want full-text, analytics and vector search under a permissive Apache 2.0 licence, with no paid feature tier in the engine itself. On AWS, Amazon OpenSearch Service and OpenSearch Serverless remove most cluster work; elsewhere you run it yourself or use a provider such as Aiven.
Overview
OpenSearch is a distributed search and analytics engine based on Elasticsearch 7.10.2, with OpenSearch Dashboards based on Kibana 7.10.2. AWS started the fork in 2021 after Elastic changed licence; in September 2024 governance moved to the OpenSearch Software Foundation, a Linux Foundation project launched with AWS, SAP and Uber as premier members. Release 3.9.0 was the latest on GitHub on 7 November 2026.
Vector search uses the knn_vector field from the k-NN plugin. The documentation lists the engines faiss (the default, with HNSW and IVF methods), lucene (HNSW and flat), jvector, and nmslib, which is now deprecated. Hybrid search (introduced in 2.11) combines keyword and semantic queries through a search pipeline. Amazon OpenSearch Service runs OpenSearch versions up to 3.5 and legacy Elasticsearch up to 7.10.
Key features
- knn_vector field with Faiss (default; HNSW and IVF), Lucene (HNSW and flat) and JVector engines; NMSLIB is deprecated
- Hybrid query with a normalization processor (score-based) or a score ranker processor using reciprocal rank fusion
- Ingest pipelines that generate embeddings with registered ML models, including connectors to externally hosted models
- Efficient k-NN filtering and post-filtering of vector results
- Disk-based vector search, memory-optimized search and vector quantization to reduce memory use
- Full-text search, aggregations and OpenSearch Dashboards in the same engine
- Apache 2.0 licence for the engine and plugins, with no paid feature tier
Integrations and ecosystem
Pros and cons
Pros
- Permissive Apache 2.0 licence with vendor-neutral Linux Foundation governance
- Choice of vector engines and methods, including IVF for large datasets
- Deep AWS integration through Amazon OpenSearch Service and Serverless
- One engine for logs, full-text and vector search
Cons
- Engine options change over time (NMSLIB deprecated), so older index mappings may need migration
- Self-managed clusters need shard, memory and node sizing expertise
- Managed AWS pricing has several dimensions (instances, storage tiers, OCUs) that make estimates harder
- Feature behaviour can differ between engines; check the engine matrix for the features you need
Pricing
OpenSearch is free under Apache 2.0, including the k-NN and neural search plugins. Managed services are paid. Amazon OpenSearch Service charges managed clusters per instance hour plus storage (AWS examples for US East: r6g.xlarge.search 0.335 USD per hour, gp3 EBS 0.122 USD per GB-month) with no minimum fee, and OpenSearch Serverless per OpenSearch Compute Unit hour (0.24 USD per OCU-hour in AWS examples) plus 0.02 USD per GB-month of storage. Other providers such as Aiven also offer managed OpenSearch.
Pricing checked on the vendor's official page on 7 November 2026. Prices change; confirm before buying.
Who it is best for
Teams that want an Apache 2.0 search engine with vector and hybrid search, especially those already on AWS who can use Amazon OpenSearch Service.
Typical use cases
- Hybrid keyword and semantic search for e-commerce or documentation sites
- RAG retrieval over large document stores on AWS
- Log analytics and observability alongside semantic search
- Similarity search over image or text embeddings with filters
Who should look elsewhere
- Applications that only need simple similarity search over modest data, where pgvector or a lightweight vector database is easier to run.
Verdict
OpenSearch is a solid choice for vector and hybrid search when an Apache 2.0 licence and neutral governance matter, and it is the natural option on AWS. Expect to manage engine and method choices carefully, and compare Elasticsearch where specific Elastic features matter more than licence terms.
Frequently asked questions
Is OpenSearch still run by AWS?
No. Since 16 September 2024 OpenSearch has been governed by the OpenSearch Software Foundation, a Linux Foundation project. AWS is a premier member and runs Amazon OpenSearch Service.
Which vector engines does OpenSearch support?
The k-NN documentation lists faiss (the default, with HNSW and IVF), lucene (HNSW and flat), jvector and nmslib. NMSLIB is marked as deprecated, so new indexes should use Faiss or Lucene.
What does Amazon OpenSearch Service cost?
Managed clusters are billed per instance hour plus EBS storage, with no minimum fee; AWS examples for US East include 0.335 USD per hour for r6g.xlarge.search. OpenSearch Serverless is billed per OCU-hour (0.24 USD in AWS examples) plus 0.02 USD per GB-month of storage.
Is OpenSearch compatible with Elasticsearch?
OpenSearch began as a fork of Elasticsearch 7.10.2, so older Elasticsearch concepts carry over, but the projects have diverged since 2021 and their vector features and APIs now differ.
Sources
- OpenSearch FAQ
- OpenSearch Software Foundation
- Linux Foundation: OpenSearch Software Foundation launch
- k-NN methods and engines
- Hybrid search documentation
- OpenSearch releases on GitHub
- What is Amazon OpenSearch Service
- Amazon OpenSearch Service pricing
Checked November 2026.
Listed in: Vector Databases & AI Search.
How we research reviews and what star ratings mean: our editorial method.