OpenSearch 2025 recap: performance and scalability improvements, search/index isolation, AI-driven features, and major community milestones.

2025 was a landmark year for OpenSearch. The project delivered its first major version in three years, achieved over one billion downloads, and positioned itself as a leading open source platform for AI powered search, observability, and analytics. Now fully under the OpenSearch Software Foundation and backed by the Linux Foundation, the ecosystem evolved rapidly to meet modern search, analytics, and AI demands.

This post highlights the key technical advancements of 2025 organized by what matters most to users: better search, faster performance, AI readiness, improved tooling, and a thriving community.

Search and Query Innovation

Hybrid and Semantic Search Enhancements

OpenSearch continued expanding hybrid search capabilities throughout 2025, blending traditional keyword search with semantic vector retrieval to improve relevance. Common filter support for hybrid queries simplified filtering across both lexical and vector components in the same query, making it easier to build sophisticated search experiences without complex workarounds.

The platform also introduced agentic search, which allows users to interact with data through natural language inputs. Agentic search leverages intelligent agents to automatically select the right tools and generate optimized queries based on user intent, eliminating the need to construct complex domain specific language (DSL) queries.

The Search Relevance Workbench provides integrated tools for teams to evaluate and optimize search quality through experimentation. Later releases added the ability to schedule experiments from the user interface, support for agentic search in the single query comparison tool, and GUID filtering for easier configuration management.

These improvements translate to higher quality search experiences with less effort. Teams can systematically measure and improve ranking strategies, while natural language querying lowers the barrier to entry for less technical users.

Performance, Scalability, and Workload Isolation

Dramatic Speed Improvements

Performance was a central theme throughout 2025. The 3.x release series delivered up to 11x faster query speeds compared to earlier versions. Hybrid search algorithms now deliver up to 65% faster query times and up to 3.5x increase in throughput.

These gains stem from multiple technical investments. The upgrade to Apache Lucene 10 improved both I/O and search parallelism. Approximation framework enhancements dramatically improved pagination performance for search_after queries. The expansion of approximate query capabilities to all numeric field types accelerated analytics workloads and time series data analysis.n

Range query performance improvements specifically benefited log analytics and time series workloads, with reduced latency for high cardinality aggregations making OpenSearch more effective for observability use cases. For large enterprise search and observability deployments, these scalability gains translate directly to better user experiences and improved productivity.

Reader and Writer Separation

One of the most significant infrastructure improvements in 2025 was the introduction of reader and writer separation, which decouples indexing and search workloads so each can scale independently.

OpenSearch introduced new shard roles to enable this separation. Write replicas act as redundant copies of the primary shard and can be promoted to primary if needed. Search replicas serve search queries exclusively and cannot become primaries. A new search node role allows operators to allocate nodes purely for serving queries.

This architecture enables dedicated indexing fleets and search fleets within the same cluster. The remote store model supports parallel segment downloads, minimizing replication overhead and reducing contention between write and read operations. Benchmarks showed roughly 30 to 40 percent improvement in write throughput for workloads with concurrent indexing and search.

With this separation, search and indexing traffic no longer compete for the same resources. Clusters can better support high write observability workloads alongside low latency search services, and operators can use specialized hardware to reduce costs while improving performance.

AI Workload Readiness

OpenSearch 3.0 introduced GPU based acceleration leveraging NVIDIA cuVS for indexing workflows. This feature accelerates index builds by up to 9.3x and can reduce costs by 3.75x, making large scale vector workloads more economically viable for production deployments.

Throughout the year, vector search received continuous enhancements. Support for vector indexing types like FP16, Byte, and Binary increased deployment flexibility. Native FP16 vector scoring further reduced compute costs for vector operations. Memory optimized search with Faiss support improved latency for vector queries, especially in high density AI workloads.

Batch processing for the semantic highlighter improved performance by reducing overhead latency and improving GPU utilization. Enhancements to the Neural Search plugin made semantic search more efficient and provided additional optimization options for specific data, performance, and relevance needs. These improvements make it practical to deploy vector search at scale for generative AI applications, retrieval augmented generation (RAG) pipelines, and semantic search experiences.

Model Context Protocol and Agentic AI

Native Model Context Protocol (MCP) support enables AI agents to communicate seamlessly with OpenSearch. This opens up more comprehensive and customizable AI powered solutions where external systems can interact with OpenSearch in standardized ways.

Agentic search and agentic memory capabilities evolved throughout the year. Agentic memory provides a persistent memory system that enables AI agents to learn, remember, and reason across conversations and interactions. This feature provides comprehensive memory management through multiple strategies including semantic fact extraction, user preference learning, and conversation summarization.

By late 2025, agentic memory reached production readiness, allowing agents to maintain context and build knowledge over time. This enables sophisticated memory operations such as consolidation, retrieval, and history tracking. The agentic search user experience now offers a redesigned, no code interface that simplifies building agents and running agentic searches. These capabilities make it easier to incorporate OpenSearch into larger AI systems where personalized, context aware responses improve over time.

Developer and Ecosystem Tooling

Protocol and Transport Improvements

OpenSearch added Protobuf and gRPC support to improve communication efficiency. The gRPC support enables faster and more efficient data transport and data processing for OpenSearch deployments, providing a new approach to data transport between clients, servers, and node to node communications.

Pull based ingestion enhances ingestion efficiency by decoupling data sources from consumers, giving OpenSearch more control over data flow and retrieval timing. This feature supports pulling data from streaming systems like Apache Kafka and Amazon Kinesis.

Arrow Flight RPC plugin additions provide server bootstrap logic and client capabilities for internode communication, further expanding the options for high performance data movement. These protocol improvements enable better cross language client support and more flexible data pipeline architectures.

Improved PPL Query Language for Analytics Enhancements

Apache Calcite became the default query engine for PPL, introducing better portability for a wide range of data management systems. This brings optimization capabilities and improvements to query processing efficiency.

PPL received significant upgrades throughout the year, with new commands including chart, streamstats, multisearch, replace, and appendpipe, along with new mvindex and mvdedup functions. PPL alerting capabilities allow users to execute monitors and view monitor statistics directly. These enhancements make complex log analytics workstreams easier and more efficient.

Index type detection enhances productivity by automatically determining whether an OpenSearch index contains log related data, speeding up log analysis feature selection.

Dashboard and UI Improvements

A completely redesigned Dashboards interface brings log analytics, distributed tracing, and intelligent visualizations into a single user experience. Users can analyze and correlate observability data with auto visualizations, context aware log and trace analysis, and a unified view across data types.

React Flow library integration provides a standardized framework for interactive node based visualizations. The library is used with Discover Traces to render service maps that visualize trace spans and service dependencies.

Experimental AI capabilities in Dashboards transform how users interact with their data through intelligent context awareness, conversational interfaces, and agent integration. The Context Provider plugin captures automatic context, while an enhanced Chat UI creates an intelligent assistant experience on the Discover page.

OpenSearch Workspaces allow administrators to assign individual accounts different dashboard views, tailoring data visualizations to each person's needs. The unified experience reduces context switching between tools and speeds up troubleshooting workflows.

Unified Insights Across Logs, Metrics, and Traces

OpenSearch expanded its observability capabilities with tight integration between logs and metrics dashboards. The Discover Traces feature provides distributed tracing visualization with service maps showing trace spans and dependencies.

Alerting with contextual insights supports trend detection and incident response. Anomaly detection insights pages make it easier to understand detected anomalies and take action.

Streaming aggregation functionality improves resource distribution by making the coordinator the single point to scale for aggregation operations.

Multi tier storage options help manage data lifecycle effectively. Amazon OpenSearch Service introduced a new multi tier storage architecture combining S3 cloud technology with local instance storage for improved durability and performance. The new warm tier supports write operations, providing greater flexibility compared to read only warm storage. This unified approach to logs, metrics, and analytics reduces toolchain complexity and provides better context for incident response.

Platform Modernization

Infrastructure and Architecture Updates

The upgrade to Apache Lucene 10 required JVM version 21 or later, bringing access to modern language features and performance improvements. Java Platform Module System support improves code organization and creates a foundation for refactoring the monolithic server module into separable libraries.

The server code structure was broken into modular pieces to increase development velocity, making the codebase easier to navigate and understand. For maintainers and contributors, this means a streamlined pull request process with changes more clearly tracked in their respective components.

Derived Source reduces storage consumption by one third by removing redundant vector data sources and utilizing primary data to recreate source documents as needed for reindexing.

Star tree optimizations now support multi terms aggregations, which are among the slowest running aggregations when applied to large datasets. These platform improvements result in a more maintainable codebase that can evolve faster, while storage cost reductions and query performance improvements translate directly to lower operating costs.

Community and Ecosystem Growth

Foundation Milestone

In August 2025, the OpenSearch Software Foundation celebrated its first anniversary under the Linux Foundation. The milestone reflected substantial growth across all dimensions.

Project downloads increased 78% year over year, bringing total downloads to more than one billion. The foundation drove community participation with more than 400 active contributing organizations and more than 8,800 contributions during its first year.

Global collaboration expanded with the highest contributions coming from the United States, Germany, the United Kingdom, Australia, and India. The membership base grew to 16 organizations with the addition of ByteDance, DataStax, DTEX, and Seacom Srl. A technical steering committee of 15 members was established, representing corporate and independent entities including Aryn, AWS, ByteDance, IBM, Paessler, Salesforce, SAP, and Uber. This growing ecosystem and vendor neutral governance provides confidence for long term investment in OpenSearch.

Leadership and Enterprise Engagement

Bianca Lewis was appointed Executive Director of the OpenSearch Software Foundation in September 2025. She brings more than two decades of experience in technology, entrepreneurship, and business strategy, having served in executive leadership positions at companies like Opster and Logz.io.

IBM joined as a Premier Member in November 2025, signaling broader corporate adoption and backing from enterprise players. IBM plans to deepen integration between OpenSearch and its open source ecosystem, improving vector search performance, multimodal document ingestion, and developer experience for AI agents.

AWS, SAP, and Uber continue as premier members and active supporters. The project also announced an Ambassador Program to recognize advocates and contributors across regions and specializations. This strong governance and enterprise backing signals that OpenSearch is production ready for demanding workloads.

OpenSearch Ambassadors Program

In 2025, the Linux Foundation launched the Ambassadors Program to recognize contributors and advocates who help grow the community, share knowledge, and drive adoption worldwide. Ambassadors support events, write tutorials, provide guidance, and serve as a bridge between the project and its users.

I’m personally honored to be one of the OpenSearch Ambassadors, and it’s been incredible to contribute to the community in this role. The program highlights the dedication of individuals helping OpenSearch thrive globally. You can learn more about the Ambassadors Program here.

Looking Ahead

OpenSearch in 2025 delivered on multiple fronts that matter to users building modern applications.

Smarter search through hybrid and semantic capabilities with improved filters and natural language querying makes it easier to build high quality search experiences.

Infrastructure isolation through reader and writer separation enables better performance and more efficient resource utilization for clusters handling both heavy indexing and demanding search workloads.

AI ready architecture with GPU acceleration, MCP support, and agentic capabilities positions OpenSearch as a foundation for generative AI applications and intelligent search experiences.

Developer friendly tooling with Protobuf, gRPC, enhanced PPL, and improved dashboards accelerates development and operations workflows.

Unified observability for logs, metrics, traces, and analytics reduces toolchain complexity and improves time to insight.

A thriving community with strong governance, enterprise backing, and global participation ensures the project will continue to evolve and improve.

Whether you are building enterprise search platforms, AI powered experiences, or observability infrastructure, OpenSearch is now better positioned than ever as an open, scalable, and extensible foundation for modern data workloads.