What We Announced at Splash 2026: A New Foundation for Finance

Every year at Splash, we use the Technical Keynote to show customers what we’ve been building — and why. This year was different. Not because of one announcement, but because of what all of them had in common: a deliberate effort to remove the constraints that have held Finance teams, developers, and data architects back.

Here’s what we shared.

A New Cube Engine: More Data, More Possibilities

For years, the reality of working with large financial models was this: more data meant more problems. Slower consolidations. Memory ceilings. Compute that buckled under pressure.

Eighteen months ago, we decided to attack that problem at every layer — storage, memory, and compute — going back to first principles on how our engine stores and accesses data on disk. The result is

OneStream’s new cube engine

a next-generation finance analytics engine that is drop-in compatible with your existing models. No migration project. No implementation changes. The engine underneath gets better; your work keeps running.

The early numbers speak for themselves:

• 2x–4x improvements across calculation, consolidation, translation, and load to cube
• One real-world consolidation cut from 1 hour 36 minutes to 28 minutes — same model, same data, same rules
• Data unit size guidance increased 10x, from 2 million to 20 million rows

The new engine introduces four innovations that unlock this scale:

1. Sparsity. Most financial data is inherently sparse — the majority of intersections have no values. Adaptive query optimization and bitmap indexes make sparsity far less of a constraint than it has ever been.
2. Native Attribute Support. Attributes are now first-class compute primitives inside the engine — not a layer on top. Query, filter, and aggregate on attributes as natively as any user-defined dimension.
3. Flexible Numeric Precision. Three new precision tiers allow implementers to configure the right level per cube — up to 4x denser storage where full precision isn’t needed.
4. Daily Granularity. Native daily-level operation, built on the new storage and memory architecture, with no performance penalty. A daily lookup is now as efficient as a monthly one.

The theme of the new cube engine is simple: more data should mean more possibilities — not more tradeoffs. That’s the foundation we’ve built.

Developer Studio: Build on Your Terms

The developer experience on enterprise Finance platforms has been constrained for too long — code locked inside a browser-based editor, no AI coding tools, no native source control. Developers became the bottleneck for everything beyond standard configuration.

OneStream Developer Studio changes that. It’s a desktop client that connects to any OneStream PV9.2.0+ environment, lets developers pull business rules and workspace assemblies as native .NET project files, and work in the IDE, source control system, and AI-assisted workflow they already rely on — then push changes back with server-side compilation and real-time feedback.

Any IDE. Any source control. Any workflow. Full AI. All on OneStream.

SensibleAI: Faster to Build With, Easier to Integrate

We’ve been building better AI capabilities. We’ve also been building better ways to work with them.

SensibleAI Studio now offers 60 reusable AI routines — the building blocks that bring AI capabilities into OneStream solutions. Developers can already orchestrate SensibleAI Forecast programmatically in C#; now Typed Clients extend that to any service that can make a REST call, making Forecast easier to integrate into analytics platforms, in-house tools, and broader enterprise software.

We also announced SensibleAI Forecast Express — a simpler, faster path to value. Cube integration is now native to the solution, meaning customers can source historical data from the cube and write forecasts directly back to it. No custom data pipelines required.

The Finance Agentic Layer: Agents Your Way

AI agents are moving fast — new clients, new model providers, new protocols every quarter. Our approach was deliberate: don’t tie customers to any single agent, model, or protocol. Build a governed layer that lets organizations bring the agents they choose.

The OneStream Finance Agentic Layer is built on three tightly integrated pillars:

• Agentic Gateway — the mandatory entry point for every agentic request, regardless of source. Universal routing, identity and access resolution at the edge, built-in rate limiting and budget controls. Every request is secure, governed, and cost-controlled.
• Agentic Toolkit — a composable surface of MCP tools, semantic layer, AI skills, and plugins. MCP tools are the verbs of the platform — purpose-built functions for searching dimensions, querying data, and executing workflows. The semantic layer keeps every agent grounded in your specific OneStream implementation and metadata. Skills package prompt engineering into reusable intelligence. Plugins bundle it all for deployment into your agents.
• Native Agents — purpose-built agents designed for Finance workflows out of the box.

Every primitive is version-controlled and regression-tested. When you build on the agentic layer, you inherit the engineering IP we’ve been building and maintaining — so your agents understand OneStream the same way ours do.

One Thread Through All of It

The new cube engine removes the ceiling on data scale. Developer Studio removes the constraints on how solutions get built. SensibleAI makes AI capabilities faster to deploy and integrate. The Finance Agentic Layer opens OneStream to the broader AI ecosystem — on your terms, with full governance.

More data. More possibilities. More freedom to build. This is the next generation of OneStream.

To learn more about everything announced at Splash 2026, visit onestream.com.