Version features

Version 9.3 (latest recommendation)

New open-source features:

• Agent Builder is generally available (GA). It allows you to interact with your Elasticsearch data in Kibana through conversations, build AI-powered Q&A applications, and use built-in agents out of the box.

• DiskBBQ allows you to search quantized vectors directly from disk without loading the full vectors into memory. This achieves a latency of less than 20 ms with only 100 MB of memory.

• The ACORN filtered vector search algorithm integrates filtering logic into HNSW graph traversal, increasing filtered search speed by up to 5x without affecting accuracy.

• Built on Lucene 10 for further improvements in index compression and inverted index retrieval efficiency.

• LOOKUP JOIN is generally available (GA). It supports cross-index join operations directly within an ES|QL query pipeline and is further enhanced to support multi-field matching, expression evaluation, and execution across remote clusters.

• ES|QL query performance for time-series data is significantly optimized, with latency reduced by up to 5x. New time-series aggregation commands such as RATE, *_OVER_TIME, TBUCKET, and TS are added.

The latest version of Alibaba Cloud Elasticsearch, based on version 9.3, reduces both memory and storage costs for vector search through BBQ and DiskBBQ. It also enhances cross-index joins and time-series analysis with ES|QL, and helps you quickly build AI search applications with Agent Builder.

For more information about the changes, see What’s new in 9.x.

Version 8.17

New open-source features:

• The dense_vector field introduces the Better Binary Quantization (BBQ) type, which compresses vector indexes by a factor of up to 32 and significantly reduces memory usage.

• Inference APIs are generally available (GA). For more information, see Inference APIs.

• Reciprocal Rank Fusion (RRF) is generally available (GA). For more information, see Reciprocal rank fusion.

• The logsdb index mode is generally available (GA). It reduces storage space for log indexes by approximately three times. For more information, see Logs data stream.

• This version introduces a built-in model for Elastic Rerank. For more information, see Elastic Rerank.

• The best_compression codec uses zstd compression. This reduces storage by approximately 12% and increases write throughput by 14%.

• Several ES|QL features are optimized, including support for full-text search. For more information, see ES|QL.

Alibaba Cloud Elasticsearch provides the latest enhanced version based on 8.17. Its built-in model services help you flexibly build AI search applications and call any external AI model service. Better Binary Quantization (BBQ) reduces memory costs by more than 10 times.

For more information about the changes, see What’s new in 8.17 and What’s new in 8.16.

Version 8.15

New open-source features:

• Vector index fields are optimized. For more information, see dense-vector.

  • By default, the int8_hnsw type replaces hnsw, and int8 quantization is enabled.

  • Support for int4 quantization is added, reducing memory usage by a factor of eight.

  • The bit vector type is added.

• SIMD instructions accelerate the merge performance of int8-quantized indexes on the aarch64 architecture, improving merge performance by a factor of approximately 3.

• This version supports the rerank phase, which allows the text_similarity_reranker to use rerank models. For more information, see text-similarity-reranker-retriever.

• This version adds the retriever query syntax to support combining results from multiple queries. For more information, see retriever.

• This version adds the semantic_text field type to better support semantic search. For more information, see semantic-text.

• The sparse_vector syntax replaces text_expansion for sparse queries. For more information, see query-dsl-sparse-vector-query.

• The query rules API is generally available (GA). For more information, see query-rules-apis.

• Index sorting supports nested fields. For more information, see index-modules-index-sorting.

• This version adds a new efficient index mode for logging scenarios, logsdb. For more information, see logs-data-stream.

• The Lucene version is upgraded to 9.11, improving memory efficiency and query performance. For more information, see apache-lucenetm-9110-available.

For more information about the changes, see What’s new in 8.15 and What’s new in 8.14.

Version 8.13

New open-source features:

• The maximum number of vector dimensions is increased to 4096. For more information, see 4096 dimension dense vector.

• Vector indexes support Scalar Quantization, which reduces the memory usage of vector indexes by nearly 75%. For more information, see scalar-quantization-in-lucene.

• This version adds the sparse_vector type, enabling support for sparse vectors. For more information, see Sparse vector.

• This version supports query parallelization for a single shard. For more information, see Query parallelization.

• This version supports the nested type for vector fields, allowing you to split a document into paragraphs and create a vector index for each paragraph. For more information, see Multiple results from the same doc with nested vectors.

• This version adds the Learning to Rank feature, which supports re-ranking results during the restore phase. For more information, see Learning To Rank.

• This version supports a new inference API for integrating external model services. For more information, see inference APIs.

• SIMD improves vector query performance. For more information, see Accelerating vector search with SIMD instructions.

For more information about the changes, see What’s new in 8.13.

Version 8.9

New open-source features:

• This version supports hybrid ranking of results from text and vector searches. For more information, see Reciprocal rank fusion (RRF).

• The maximum number of vector dimensions is increased to 2048. For more information, see Increase max number of vector dims to 2048.

• This version improves brute-force search performance. For more information, see Improve brute force vector search speed.

• kNN queries support searching across multiple fields simultaneously. For more information, see Allow more than one KNN search clause.

• The built-in ELSER model is available. For more information, see ELSER inference integration.

• Stable support for distributed NLP model scheduling and management is available. For more information, see Make native inference generally available.

• This version improves write performance for documents with primary keys. For more information, see Primary key optimization.

• This version improves query performance for constant keyword fields. For more information, see Skip shards when querying constant keyword fields.

• Time Series Data Streams (TSDS) and the downsampling feature are stable. For more information, see TSDS and Downsample.

• This version optimizes memory usage for source documents by removing the use of ThreadLocal. For more information, see Remove uses of deprecated LeafReader.

For more information about the changes, see What’s new in 8.9.

Version 8.5

New open-source features:

• This version implements vector similarity search based on the HNSW algorithm. For more information, see k-nearest neighbor (kNN) search.

• This version adds the Time Series Data Stream (TSDS) feature. For more information, see Time series data stream (TSDS).

• This version adds geo grid queries. For more information, see Geo grid query.

• Security configurations are simplified. For more information, see Start the Elastic Stack with security enabled automatically.

• This version improves the Lucene compression algorithm to reduce index size.

• This version enhances range query performance.

• This version supports the lookup runtime field type. For more information, see lookup-runtime-fields.

• This version implements random sampler aggregation queries. For more information, see Random sampler aggregation.

• The heap memory consumption of master and data nodes is reduced.

• The _type mapping is removed. However, version 8.x is backward compatible with requests from version 7.x. For more information about the compatibility mode, see rest-api-compatibility.

• Index protection is implemented. By default, the elastic user can only read built-in Elasticsearch indexes.

For more information about the changes, see Breaking changes in 8.5.

Version 7.16

New open-source features:

• This version supports SQL queries for cross-cluster search.

• Ingest pipelines support enrich policies for the range type.

• This version optimizes caching to improve query performance.

• This version adds support for adding and removing an index from a data stream.

• This version adds the cluster UUID and name to audit logs.

For more information about the changes, see breaking changes in 7.16.

Version 7.10

New open-source features:

• This version improves the compression of stored fields to reduce storage costs.

• Event Query Language (EQL) enhances Elasticsearch security. For more information, see Event Query Language (EQL).

• The default value of search.max_buckets is increased from 10,000 to 65,535.

• This version supports case-insensitive queries. You can set the optional case_insensitive parameter to true to enable this feature.

For more information about the changes, see Breaking changes in 7.10.

Version 7.7

New open-source features:

• When you create an index, the default number of primary shards changes from 5 to 1.

• Mapping types are removed. You no longer need to specify a type when you define an index mapping or template. For more information, see Removal of mapping types.

• By default, a search request returns a maximum of 10,000 documents. If a query matches more than 10,000 documents, only the first 10,000 are returned. For more information, see track_total_hits 10000 default.

• A single data node can have a maximum of 1,000 shards by default. You can configure this limit by using the cluster.max_shards_per_node parameter. For more information, see Cluster Shard Limit.

• A maximum of 500 scroll contexts can be open at one time by default. You can configure this limit by using the search.max_open_scroll_context parameter. For more information, see Scroll Search Context.

• The parent circuit breaker now considers the actual memory available (indices.breaker.total.use_real_memory), which defaults to 95% of the JVM heap memory. This helps prevent out of memory (OOM) errors by using the maximum available memory more effectively. For more information, see Circuit Breaker.

• Support for the _all field is deprecated to improve retrieval performance.

• This version adds intervals queries. They allow you to search based on the order and proximity of multiple terms in a text. For more information, see Intervals Queries.

• When audit logging is enabled, audit events are persisted to the <clustername>_audit.json file on the host's file system. Storing audit events in an index is not supported. For more information, see Enabling audit logging.

For more information about the changes, see Breaking changes in 7.0.

6.x versions (6.7, 6.8)

New open-source features:

• An index can have only one type. The recommended type is _doc.

• Version 6.6.0 adds index lifecycle management (ILM) to reduce index maintenance costs.

• This version adds the Rolling up historical data feature to summarize historical data.

• Version 6.3 adds support for X-Pack SQL, which allows you to convert SQL queries into DSL queries and reduces the effort required to learn DSL.

• Aggregation functions are enhanced with support for Composite, Parent, and Weighted Avg.

For more information about the changes, see Breaking changes in 6.0.

5.x versions (5.6)

New open-source features:

• An index can have multiple types, and custom types are supported.

• The string field type is deprecated and replaced by text or keyword.

• The values for the index field change from not_analyzed or no to true or false.

• This version uses float instead of double to reduce storage costs.

• The Java High Level REST Client is introduced to replace the TransportClient.

For more information about the changes, see Breaking changes in 5.0.