OceanBase Named a Strong Performer in The Forrester Wave™: Multimodel Data Platforms

AI are changing what enterprises expect from data infrastructure.

Instead of serving only applications, databases increasingly need to provide continuous, trusted context for autonomous systems operating in real time. This shift is also reshaping how analysts evaluate modern data platforms.

Today, OceanBase is proud to be recognized as a Strong Performer in The Forrester Wave™: Multimodel Data Platforms, Q2 2026.

As AI agents transition from conversational interfaces into mission-critical production workflows, they demand continuous, real-time access to memory, context, state, and enterprise data. Simultaneously, enterprises have adopted specialized engines — relational for transactions, document for flexibility, graph for relationships, vector for retrieval, and time-series for telemetry — gaining best-of-breed capabilities at the cost of operational fragmentation.

Data movement between these systems through pipelines introduces latency, consistency gaps, and governance seams. Consequently, traditional architectures struggle to deliver the real-time, consistent, and governed context that AI agents require. Multimodel Data Platforms (MMDPs) address this complexity by supporting multiple data models within a unified engine, with shared transaction semantics, query interfaces, governance, and operational tooling.

So far, OceanBase addresses these challenges through a single-engine multimodel architecture. It provides hybrid search and shared storage to enable concurrent access and scalable deployments.

Inside the Multimodel Data Infra

Behind that unified architecture is a specific design choice: structured business data and multimodal data share one foundation — one metadata system, one permissions model, one lifecycle, and one query path.

• Multimodal tables. A single table semantic describes richer business entities that include structured fields, JSON, documents, images, audio, video, large objects, vectors, and model-generated results. A customer service case can hold ticket fields, chat history, call recordings, uploaded images, and embeddings in one row. A logistics record can hold order metadata, route events, driver notes, delivery images, and semantic features together. Internally, different data types use different storage layouts; externally, they are managed as one.
• AI columns. Model-generated outputs — embeddings, summaries, labels, extracted features — are materialized as columns inside the data system rather than assembled through external pipelines. When source data is written, embedding and tagging can be triggered and committed transactionally with the row. AI-generated results become governed data assets, not hidden pipeline artifacts.
• Hybrid search. Keyword search, vector search, and structured filtering run in one query path, in one engine, over one data copy — no external index syncing, no drift between the search layer and the source of truth. For agents that act on retrieved context, this matters: stale results are not just quality issues; they are production risks.
• Shared storage, open compute. A single physical copy of data simultaneously supports operational transactions, real-time analytics, AI retrieval, and open compute engines such as Spark and Ray — no export, no copy, no cross-system pipelines. Storage and compute are disaggregated so each scales independently, and S3-compatible object storage with Iceberg-format tables reduces lock-in while preserving the transactions, consistency, and governance that production AI depends on.

Together, these capabilities make it possible for enterprises to deliver real-time, trusted context to AI agents on top of the same platform running their transactional systems — rather than building a parallel data stack for AI.

The Forrester Wave™ report evaluated 14 vendors from various geographies across 24 criteria in two categories: Current Offering and Strategy. We achieved the highest possible score of 5 in five criteria, including Multimodel Transactional Consistency, Multimodel Translytical Capabilities, and Deployment Flexibility (Cloud, Edge, On-Premise, and Hybrid).

"OceanBase's strategy envisions its platform operating as a single coherent system in which every data model is a first-class citizen," stated the Forrester report. "(It) is well-suited to enterprises that require a resilient, single-engine distributed multimodel platform to simultaneously process high-concurrency transactions and run real-time analytics without latency."

This vision also shapes the evolution of OceanBase AI Database. Announced in June 2026, it extends our multimodel architecture into a broader AI data infrastructure, helping enterprises manage multimodal data, provide trusted real-time context for AI agents, and reduce the complexity of fragmented data stacks.