We are thrilled to introduce Oracle AI Database 26ai, our next-generation AI-native database, with AI architected into the entire data and development stack. Oracle AI Database enables you to easily deliver trusted AI-powered insights, innovations, and productivity for all your data everywhere, including both operational systems and analytic data lakes. You can now run dynamic agentic AI workflows to provide sophisticated answers and actions that combine private database data with public information.
Oracle AI Database 26ai replaces Oracle Database 23ai. Transitioning from 23ai to 26ai is simple—just apply the October 2025 release update with no database upgrade or application re-certification. Advanced AI features like AI Vector Search are included at no additional charge.
Since the introduction of Oracle Database 23ai, we’ve delivered a steady cadence of AI-first capabilities that broaden functionality and increase performance. Those updates:
- extended AI Vector search capabilities,
- expanded support for combining vector search with all other data types,
- added support for agentic AI, and
- hardened data privacy governance and observability
paving the way for Oracle AI Database 26ai’s unified AI experience. In short, 23ai established the foundation; 26ai consolidates and elevates it into a cohesive, AI-native platform for production workloads at scale.
Let’s take a tour through highlights of capabilities that will be available in Oracle AI Database 26ai, which we’ll refer to as AI Database in the rest of this post.
For retrieval and reasoning using multimodal enterprise data, such as PDFs, videos, and images, AI Database introduces Unified Hybrid Vector Search. You can blend vectors with relational, text, JSON, graph, and spatial predicates in a single query to retrieve documents, images, audio, video, and table rows together. Pair that with industry-leading LLMs and the Model Context Protocol (MCP) to enable agentic workflows that iterate, fetch additional context, and deliver higher-quality answers—grounded in your private data.
We’ve made agents first-class citizens in the database. Select AI Agent lets you define, run, and govern AI agents inside Autonomous AI Database using in-database tools, external tools over REST, or MCP Servers. For teams that prefer a no-code approach, the Private Agent Factory packages a builder and deployment framework you can run in any environment you control, so you keep data private while benefiting from the full performance and security of the database.
You have flexible model choices without losing control of data. Use ONNX embedding models and integrate with LLM providers or run private inference via Private AI Services Container to avoid sending data to third-party AI services. AI Database also supports integration with NVIDIA NIM containers for embeddings and RAG pipelines, and its design accommodates future GPU acceleration for vector indexing with NVIDIA CAGRA and cuVS.
AI Database expands your analytics surface area without fragmenting it. The new Autonomous AI Lakehouse transforms your warehouse or data lake into a truly open and multi-vendor AI lake house. Autonomous AI Lakehouse supports reading and writing Apache Iceberg data formats in the object store, so you can bring all the benefits of Oracle Analytics such as Exadata-powered performance and serverless scaling to open lakehouse tables across OCI, AWS, Azure, and Google Cloud. Autonomous AI Lakehouse is interoperable with Databricks, Snowflake, and many other Iceberg compliant data stores on the same clouds allowing seamless data sharing between data stores without moving the data.
With Oracle AI database, Developers get a simpler, more powerful method of working with data. Relational, JSON, and graph are unified, so the same data can be accessed via SQL, as documents, or as graphs—no brittle ETL hops. Data Annotations let you describe purpose, semantics, and constraints, so AI-powered tooling and agents are prepared to generate better code and more accurate answers. For rapid app delivery, the upcoming APEX AI Application Generator will use natural language to build enterprise-grade experiences faster.
Performance and elasticity are engineered into the stack. Oracle Exadata for AI offloads vector search to intelligent storage for dramatic speedups and scales both to high-end workloads and, with the new Exascale software architecture, to smaller, lower-cost deployments—ideal for smaller enterprises or workgroups, or teams piloting AI features and then ramping to production. True Cache brings an application-transparent, transactionally consistent mid-tier cache, so you can keep latency low while retaining full SQL compatibility including Oracle Relational, Vector, JSON, Spatial, and Graph query semantics.
Security and data governance are enforced in-database. Built-in Data Privacy Protection will enable row-, column-, and cell-level controls plus dynamic data masking, so both users and AI agents only see authorized data. SQL Firewall adds scalable, in-database protection against unauthorized SQL and injection attacks without adding a mid-tier server that adds extra network latency and management costs. For crypto-agility, AI Database implements NIST-approved quantum-resistant algorithms (ML-KEM) for data-in-flight and already supports quantum-resistant encryption for data-at-rest.
Mission-critical operations are covered from end to end. Oracle Database Zero Data Loss Cloud Protect provides on-premises databases with real-time protection by continuously shipping changed data to the cloud — helping defend against data loss and ransomware. Globally Distributed Database delivers multi-master, active-active deployments with built-in RAFT replication and sub-three-second failover with zero data loss, supporting both ultra-scale and data sovereignty requirements.
Bottom line: Oracle AI Database uniquely brings AI to both mission-critical OLTP systems and data lakes, allowing you to easily build trustworthy, high performance AI solutions on the databases that already run your business. It enables delivery of trusted AI-powered insights, innovations, and productivity for all data, across the cloud, multicloud, and on-premises. New AI capabilities are available everywhere, for no additional charge.
Oracle has adopted a cloud-first, developer-first strategy. Pursuant to that strategy we have released Oracle AI Database first on the following cloud platforms:
- Oracle Autonomous Database
- Oracle Exadata Database Service
- Oracle Base Database Service
- Oracle Database@Azure
- Oracle Database@Google Cloud
- Oracle Database@AWS
- Oracle Database installed in Oracle Cloud Infrastructure Compute VMs
For on-premises, we have released Oracle AI Database on the following platforms:
- Exadata Cloud@Customer
- Compute Cloud@Customer
- Oracle Exadata Database Machine
- Oracle Database Appliance
- Oracle Private Cloud Appliance
To quickly and easily try AI Database you can simply go to FreeSQL.com. For development and testing, we have released Oracle AI Database Free for Linux x86, ARM Linux, and Windows. For larger scale development and testing, we suggest using cloud services such as Oracle Autonomous AI Database, Oracle Exadata Exascale, Oracle Base Database, running your own Oracle AI Database on Oracle Cloud Infrastructure, or enrolling in our Beta program.
While the date for Oracle AI Database on other platforms has not been announced, we will update My Oracle Support (MOS) Doc ID: 742060.1 when it is announced. Please continue to refer to this MOS document for all Oracle Database release and support schedules.