Oracle brings AI model access to 30 more cloud regions and adds agent workspaces

Oracle is giving customers access to AI models in more cloud regions without requiring those models to run locally. On October 26, 2026, it introduced Smart Model Router and Agent Sandbox for OCI Enterprise AI, pairing access across 30 additional commercial regions with generally available, temporary Linux workspaces where agents can execute tasks.

The model can run somewhere else

Smart Model Router addresses a location problem: model availability and GPU capacity vary across cloud regions. Oracle says developers otherwise face managing regional model endpoints and building their own routing logic. Its new service handles the destination choice within boundaries customers set.

Customers create reusable routing profiles that specify a supported model and the regions permitted to process its requests. OCI then routes inference—the work of producing a model response—to an approved GPU-backed region with available capacity. The customer defines the eligible destinations; Oracle manages routing within that set.

The regional expansion covers supported on-demand models and selected Enterprise AI features, including hosted deployments, sandboxes and API keys. It does not extend every model or feature to every location.

A temporary computer for the agent's work

Agent Sandbox handles a different part of an AI application: executing tasks rather than obtaining model responses. Now generally available, it provides a managed environment where an agent can create temporary Linux workspaces, perform work, collect the results and stop the workspaces.

Those workspaces can test generated code, run shell commands, analyze business data or execute diagnostic scripts. The workspace is therefore a place for the agent to carry out computational tasks, not simply return instructions for someone else to follow. Oracle operates the infrastructure underneath those tasks.

Organizations configure the compute resources, execution duration and network access. That division leaves customers responsible for choosing the operating boundaries while OCI manages the execution platform. Oracle's stated benefit is that teams can provide a controlled place for agent tasks without building and maintaining a separate platform themselves.

Where Oracle expects the tools to fit

Oracle presents the two capabilities as infrastructure for global AI applications and business automation. Its proposed uses include:

• Customer and employee applications: extend AI-powered experiences to more regions, using model capacity in approved OCI locations without requiring local GPU infrastructure.
• Business analytics: let agents analyze data and execute code to produce reports or visualizations inside controlled workspaces.
• Workflow automation: give agents an environment to process files, use approved tools and carry out multistep business tasks within configured execution boundaries.