You can use a Microsoft Foundry model with inference processed beyond the Foundry resource’s region through a global deployment, a data-zone deployment, or a regional deployment in another supported region. The right option depends on where processing is permitted, what capacity you need, and whether the exact model supports that deployment type and region. These options have different processing boundaries; the Foundry resource’s region alone does not define them.
How do I use an Azure OpenAI model in another region?
In Microsoft Foundry, choose a deployment type whose routing scope matches your organization’s requirements. “Another region” can mean Azure selects an available region globally, processing stays within a defined data zone, or you deploy the model in a particular supported region. These are not equivalent residency choices.
As an Amazon Associate I earn from qualifying purchases.
Microsoft says data at rest remains in the designated Azure geography, while inference processing depends on the deployment type. Global deployments can process prompts and responses in any Azure region where the model is deployed; data-zone deployments limit processing to the applicable Microsoft-defined zone; regional deployments process in the deployment region. Check the current Microsoft deployment types documentation and model availability table for the exact model, region, deployment type, and cloud.
Four ways to route or place a model
1. Global Standard: let Azure route inference across regions
Global Standard is the general-purpose option when your organization permits broad cross-region processing. It is pay-per-token, and Azure dynamically routes requests to available datacenters. Microsoft’s guidance says prompts and responses may be processed in any Azure region where the model is deployed. This is not a single-region residency option. See Global Standard deployment guidance.
#1 Best Overall
2. Global Provisioned: reserve capacity with global routing
Global Provisioned retains global routing while reserving throughput. Consider it when a workload needs dedicated capacity and more predictable throughput behavior than a standard deployment. The routing scope remains global, so reserving capacity does not make processing region-specific. Review provisioned deployment guidance.
3. Data Zone Standard or Data Zone Provisioned: keep processing within a defined zone
Use a data-zone deployment when processing may cross regional boundaries but must stay within the applicable Microsoft-defined zone. Data Zone Standard is pay-per-token; Data Zone Provisioned reserves throughput. A zone is broader than one region, so neither option should be treated as a guarantee of single-region processing. Confirm the zone definition and model availability in Microsoft’s current deployment types documentation and data-zone availability guidance.
4. Deploy regionally elsewhere, or consider Model Router where supported
If processing must occur in a particular region, create a regional deployment there—but only if that model and deployment type are supported in the target region. Regional Standard is pay-per-token; Regional Provisioned reserves capacity in the deployment region. Verify availability in the live model and region table.
Model Router is a separate option, not a synonym for a regional deployment. Its supported regions and underlying models are limited by Microsoft’s current support matrix; it is not an unrestricted proxy to any model in any region. Check the Model Router documentation.
Rank #3
How to choose: geography, billing, capacity, and variability
| Option | Inference processing scope | Capacity and billing | Important limitation |
|---|---|---|---|
| Global Standard | Any Azure region where the model is deployed, per Microsoft deployment guidance. | Pay per token. | Broadest processing scope; latency variability may increase at high, consistent volume. |
| Global Provisioned | Global cross-region routing. | Reserved throughput. | Capacity reservation does not narrow the processing scope. |
| Data Zone Standard | Within the applicable Microsoft-defined data zone. | Pay per token. | A zone is not a single region; confirm the applicable zone and model support. |
| Data Zone Provisioned | Within the applicable Microsoft-defined data zone. | Reserved throughput. | A zone is not a single region; confirm the applicable zone and model support. |
| Regional Standard | The deployment region. | Pay per token. | Available only for supported model and region combinations. |
| Regional Provisioned | The deployment region. | Reserved throughput in that region. | Available only for supported model and region combinations; capacity is tied to the deployment region. |
Standard deployments bill by token; provisioned deployments reserve throughput. Microsoft notes that global standard may have increased latency variability at high, consistent volume, while provisioned options provide dedicated capacity and more predictable throughput. These distinctions do not establish a specific latency for your application: measure against your own workload and deployment. Batch deployments are for asynchronous work, not a real-time substitute. Microsoft describes Global Batch as 50% lower cost than Global Standard with a 24-hour target turnaround; that is a stated product term, and a batch request can take longer than the target. See batch deployment guidance.
Check support before choosing a destination
- Identify your boundary. Decide whether processing may happen across Azure regions, must stay within a specified data zone, or must be confined to a particular region.
- Find the exact model and deployment type. Consult Microsoft’s current model availability table; support can differ by model, deployment type, target region, and Azure cloud environment.
- Compare capacity needs. Choose standard for pay-per-token usage or provisioned when reserved throughput is appropriate. Treat batch as an asynchronous workload option.
- Create the deployment in Foundry. Select the verified model, deployment type, and supported region or zone. A regional deployment’s location is the deployment region; global and data-zone deployments have broader processing scopes than the resource’s region.
- Validate governance and workload behavior. Confirm that the selected processing scope meets your organization’s requirements, then evaluate throughput and latency with your application’s traffic pattern.
Can Microsoft Foundry route inference to another Azure region?
Yes. Global Standard and Global Provisioned use cross-region routing; Data Zone types use cross-region routing within a defined zone; a regional deployment can run in another region when the model and deployment type are available there. The critical distinction is the permitted inference geography, not merely the region in which the Foundry resource exists. Availability and processing rules can change, and sovereign or government clouds may have different support matrices, so use Microsoft’s current tables for the environment you operate.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
