Plan & manage an Azure AI solution ยท 25โ€“30%

Set up solutions & deployment options

~2 min read

A deployment is a named, callable instance of a model in your Foundry resource โ€” your code targets it by deployment name. The deployment type decides region, billing, and throughput.

Deployment types โ€” memorize this table

TypeRegionBillingUse when
Global StandardAnyPay-per-tokenDefault choice. Highest quota.
Global ProvisionedAnyReserved PTUPredictable high throughput
Global BatchAny50% discount, โ‰ค24h asyncLarge async jobs
Data Zone StandardEU/US zonePay-per-tokenData-zone compliance
Data Zone ProvisionedEU/US zonePTUCompliance + throughput
Standard (regional)Single regionPay-per-tokenRegional data residency, low volume
Regional ProvisionedSingle regionPTUResidency + throughput
DeveloperAnyPay-per-tokenFine-tuned model evaluation only
PTU = provisioned throughput unit. Exam pattern: “predictable throughput” means Provisioned; “cheapest for an overnight bulk job” means Global Batch; “EU data must stay in the EU” means Data Zone; “single-region residency” means Standard/Regional.

Connect an app โ€” 3 things you always need

  1. Endpoint URL โ€” project endpoint (Foundry SDK features) or OpenAI-v1 endpoint (OpenAI-compatible).
  2. Auth โ€” API key, or Microsoft Entra ID token (recommended for production).
  3. Deployment name โ€” goes in the model parameter of requests.
from azure.ai.projects import AIProjectClient
from azure.identity import DefaultAzureCredential

project = AIProjectClient(endpoint=PROJECT_ENDPOINT, credential=DefaultAzureCredential())
client = project.get_openai_client()
response = client.responses.create(model="my-deployment", input=[...])

Dev tooling

Older classic Foundry projects use hub-based architecture. New projects: Foundry resource + projects. If a question contrasts them, “hub” = classic.