Home › Certifications › Azure AI Fundamentals AI-901 › Microsoft Foundry — Models, Deployment and Prompts
48 multiple-choice questions and 19 flashcards on Microsoft Foundry — Models, Deployment and Prompts, about 19% of the Azure AI Fundamentals AI-901 bank. Every one carries a written rationale.
Microsoft Foundry — Models, Deployment and Prompts is one of 6 chapters in CoStudy's Azure AI Fundamentals (AI-901) bank, and it holds 48 of the bank's 250 multiple-choice questions — roughly 19% of the total. That proportion is not arbitrary: chapters follow the certifying body's published exam outline, and the number of questions in each is set by that domain's published weight, so the share of your practice time this chapter takes matches the share of the real exam it accounts for.
Studying by chapter is worth doing once you have a diagnostic score. A single overall percentage tells you whether you are close; it does not tell you which domain is dragging. Working a weak chapter in isolation, and re-testing it in isolation, is the fastest way to move a score that has stalled — and it is why the mock exams in CoStudy report by domain rather than as one number.
5 questions drawn from this chapter, with the full rationale shown — the controlling principle behind the right answer, and why each wrong option tempts and fails.
A team's outputs are inconsistently formatted despite a clear system prompt describing the desired JSON shape. The change MOST likely to improve consistency is:
Answer: C — Adding two or three worked input-output examples to the prompt
C) Correct — few-shot examples show the exact target shape, which constrains formatting far more reliably than description alone. A) Higher temperature increases variability, worsening the symptom. B) Deployment names have no effect on output. D) A second deployment is a capacity and latency measure, not a formatting one.
A support tool must transcribe recorded calls and then summarize them. The MOST appropriate approach in Microsoft Foundry is:
Answer: B — Use Azure Speech in Foundry Tools, then a generative model to summarize
B) Correct — speech-to-text produces the transcript, and a generative model then does the language task over that text. A) Image generation cannot process audio. C) Translation converts between languages; it does not transcribe or summarize. D) Fine-tuning on raw waveforms is not how a text model is adapted.
Azure OpenAI in Foundry Models gives you:
Answer: B — Managed access to OpenAI models with enterprise controls
B) Right answer - the service runs OpenAI models inside Azure with enterprise identity, networking and regional data handling applied. A) These are proprietary OpenAI models; open-weight models sit elsewhere in the model catalog. C) Usage is billed by consumption or reserved capacity. D) The hardware is abstracted away from the customer.
An analyst is told that a prompt, the conversation history and the generated reply all share a single budget. That budget is the:
Answer: C — Model context window
C) Correct — input and output tokens together must fit inside the model's context window, which is why long histories crowd out the answer. A) Quota governs throughput across deployments, not a single call's size. B) Severity scales classify harm levels. D) Connection limits concern linked resources.
Which best describes Microsoft Foundry's relationship to Azure OpenAI in Foundry Models?
Answer: A — Foundry is the portal used to build on them
A) Correct — Microsoft Foundry is the build, deploy and evaluate hub that sits over the models plus the wider Foundry Tools capabilities. B) They are complementary layers. C) The models are still there, reached through Foundry. D) Azure DevOps is a different stack.
4 cards from the 19 in this chapter.
Microsoft Foundry?
The unified platform and portal for building AI applications and agents. • Brings together the model catalog and deployments, prompts, grounding data, content safety, evaluations and tracing in one project.
What is a connection in a Foundry project?
A stored, named link to an external resource the project needs — a search index, a storage account, another model endpoint — holding the credential centrally so agents and code reference the connection instead of embedding secrets.
What are the standard structural parts of a well-built prompt?
Instruction — what to do • Context or grounding content to use • Examples of the desired behaviour • The input to act on • A statement of the required output format. Separating these clearly, and putting the instruction near the content it governs, measurably improves reliability.
Why should the system prompt state what the assistant must NOT do?
Because a model will attempt almost anything asked of it. Explicit boundaries — no legal or medical advice, no discussing competitors, no promising refunds, escalate instead — are what stop confident out-of-scope answers, and they are far more reliable than hoping the model infers the limits.
These are a sample. The full Microsoft Foundry — Models, Deployment and Prompts chapter runs 67 items with per-chapter progress tracking, on the web and in the iOS app.
Open Azure AI Fundamentals AI-901 in CoStudy →