Azure OpenAI Service Deployment Types

Deployment Type	Description	Notes
Global	Inference requests could be routed to any data-center in the world where the LLM is supported, depending on capacity available.	Least expensive; however, it may not meet customer compliance and latency requirements.
Data Zone	Inference requests could be routed to any data center within a Microsoft-defined Data Zone.	Costs 10% more than Global; currently, there are only U.S. and Europe Data Zones.
Regional	Inference requests are routed to the Azure region where the customer has deployed the associated Open AI resource.	Twice the cost of Global for hourly pricing under the Provisioned purchase model.

The Azure OpenAI service offers three deployment type choices. Customers can choose the most cost-effective type that meets their compliance and connectivity latency requirements. The types vary by the Azure datacenters that respond to service inference requests. 

The Global type routes requests to any datacenter in the world where the required large language model (LLM) is supported, based on available capacity. Global is the least-expensive deployment type, but it provides no customer control over where request processing occurs.