Question

Your team runs a customer-facing chat application backed by a GPT-4o model deployed in Azure AI Foundry using a standard (pay-as-you-go) deployment. During a marketing campaign, traffic tripled and users began receiving HTTP 429 responses. Leadership requires predictable, guaranteed latency for the model during future campaigns, and the workload has a steady, high-volume baseline. Which action best meets these requirements?

Accepted Answer

Create a Provisioned Throughput Unit (PTU) deployment for the model to reserve dedicated capacity with predictable performance.

Answer

Implement client-side exponential backoff and retry logic against the existing standard deployment.

Answer

Request a Tokens-Per-Minute (TPM) quota increase for the standard deployment through a support ticket.

Answer

Deploy the same model to multiple regions and use a load balancer to distribute standard-deployment traffic.