Azure OpenAI Service Purchase Models

Purchase Model	Description	Capacity	Consumption	Cost
Standard	Pay-as-you-go (PAYG), based on input and output token consumption.		X	Highest.
Provisioned Hourly	PAYG, no reservation required, based on Provisioned Throughput Unit (PTU) usage.	X		Typically less than Standard.
Provisioned Monthly	Upfront capacity purchase, based on PTU usage.	X		About 67%–80% less than Provisioned Hourly.
Provisioned Annual	Upfront capacity purchase, based on PTU usage.	X		About 20% less than Provisioned Monthly.
Batch	PAYG, based on input and output token consumption. Customer provides a large number of requests in a single request and a response is generated within 24 hours.		X	About 50% less than Standard.

The Azure OpenAI service offers multiple purchase models. Customers can optimize costs by choosing the most cost-effective purchase model that meets their requirements.

The Standard (consumption-based, pay-as-you-go) purchase model charges based on the number of tokens (chunks of text) that it processes for inputs (called prompts) and outputs (the responses generated by the service).