Also known as: PT
Provisioned Throughput (PT) is a reserved-capacity service for Vertex AI generative models that provides capacity assurance and predictable pricing. Instead of relying on shared on-demand capacity, customers purchase dedicated throughput for specified models, regions, and subscription periods.
Provisioned Throughput (PT) is a premium service for Vertex AI's generative AI models that offers a guaranteed experience through capacity assurance and predictable pricing. Unlike the pay-as-you-go (on-demand) option, PT lets customers purchase a dedicated quota, ensuring their requests don't compete with others for model capacity. PT is a fixed-cost monthly or weekly subscription that reserves throughput for specified models and locations For more information, see Provisioned Throughput overview.