Cut Azure OpenAI Costs with Reservations

• The Azure OpenAI service, which can be used to build applications driven by generative AI, is offered through multiple purchase models.
• The Provisioned purchase model provides guaranteed capacity for an upfront monthly or annual customer commitment (reservation).
• The Provisioned model could cut costs compared to the Standard model, but estimating the capacity needed for new projects may be difficult without production usage data.
• Recently introduced deployment types that route requests to particular geographical areas also offer potential cost savings.

Updates to the Azure OpenAI service purchase models and new cost-cutting deployment types were introduced in the second half of 2024. This report reflects these updates, although the older Commitment purchase model also remains available for now. For details on the Commitment purchase model, see “Provisioned Throughput Could Cut Azure OpenAI Costs” on page 24 of the July 2024 Update.

(For a chart that lists and compares the purchase models, see “Azure OpenAI Service Purchase Models.” For a chart that lists and compares the deployment types, see “Azure OpenAI Service Deployment Types.”)