Choosing a Deployment Type: 10 practice questions

Question 1 of 10

A team with no data residency or throughput constraints deploys a newly released model. Which deployment type is the recommended starting point?

1. ARegional Provisioned
2. BDeveloper deployment
3. CGlobal Standard
4. DData Zone Standard

Question 2 of 10

Which deployment type is meant only for evaluating a fine-tuned model, has no SLA, and is deleted automatically after 24 hours?

1. AGlobal Standard
2. BData Zone Batch
3. CGlobal Batch
4. DDeveloper

Question 3 of 10

Compared with Global Standard, what does Global Batch offer for large asynchronous jobs?

1. AProcessing that stays inside one Azure region
2. BLower latency variance through reserved PTUs
3. CA guaranteed 24-hour completion under an SLA
4. D50% lower cost and a 24-hour target turnaround

Question 4 of 10

A European company must keep inference processing inside the EU data zone. Its traffic is bursty and it wants pay-per-token billing. Which deployment type fits?

1. AGlobal Standard
2. BData Zone Provisioned
3. CGlobal Provisioned
4. DData Zone Standard

Question 5 of 10

A subscription has enough Data Zone Provisioned PTU quota in a region, yet creating a provisioned deployment there fails. What is the most likely cause?

1. AThe model needs its own model-specific PTU quota
2. BThe PTU quota first requires an Azure Reservation
3. CProvisioned types are billed per token, not per PTU
4. DThe region lacks available capacity

Question 6 of 10

A team wants the discounted reservation rate for a new provisioned deployment. In which order should it act?

1. ABuy the reservation, then request PTU capacity
2. BBuy the reservation to secure capacity, then deploy
3. CRequest more PTU quota, then buy the reservation
4. DCreate the deployment, then buy the reservation

Question 7 of 10

A team deploys an open-source Qwen model with managed compute. They plan to use the GPU quota they already hold for Azure virtual machines. What should they expect?

1. ATheir Azure VM GPU quota covers the deployment
2. BReserved PTU quota shared with provisioned types
3. CA separate Foundry quota per accelerator SKU
4. DPay-per-token billing like Global Standard

Question 8 of 10

A customer chat service has steady, predictable high volume and needs consistent low latency. Processing must stay within the EU data zone, and the team wants higher availability than a single-region deployment gives. Which deployment type fits best?

1. AData Zone Standard
2. BData Zone Provisioned
3. CRegional Provisioned
4. DGlobal Provisioned

Question 9 of 10

Each night, 50,000 prompts must be processed: half by one model deployment and half by another. The team accepts a 24-hour target rather than real-time results, cost matters, and the online chat quota must not be affected. What setup fits?

1. AGlobal Standard calls spread evenly over the night
2. BTwo Global Batch jobs, one input file per deployment
3. CA Developer deployment for each of the two models
4. DOne Global Batch job whose file lists both deployments

Question 10 of 10

A production workload needs a partner model, built-in content filtering, an SLA, and inference processing inside a data zone. The model is offered both as a Serverless API deployment and on managed compute. Which option fits?

1. AA Serverless API deployment with a Data Zone type
2. BInstant access, calling the model by its name
3. CA managed compute deployment in the Foundry project
4. DA managed compute deployment with Data Zone routing