You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
description: "This unit provides a knowledge check on Microsoft Foundry's generative AI models and agents."
7
+
ms.date: 1/28/2026
8
+
author: sherzyang
9
+
ms.author: sheryang
10
+
ms.topic: unit
11
+
durationInMinutes: 5
12
+
content: |
13
+
quiz:
14
+
questions:
15
+
- content: "What best describes Foundry's model catalog?"
16
+
choices:
17
+
- content: "A catalog consisting of only Microsoft-exclusive foundation models"
18
+
isCorrect: false
19
+
explanation: "Foundry's model catalog includes models from many sources, not only Microsoft. It contains Microsoft‑hosted models as well as models from partners, open‑source communities, and third‑party providers."
20
+
- content: "A central hub to discover, filter, compare, and test many generative AI models from multiple providers"
21
+
isCorrect: true
22
+
explanation: "Foundry's model catalog is described as a central hub where you can browse, filter, compare, and evaluate a wide variety of generative AI models."
23
+
- content: "A tool that replaces the need for an Azure subscription"
24
+
isCorrect: false
25
+
explanation: "An Azure subscription is required to access and use Foundry. Foundry projects, deployments, model usage, and governance features are all tied to Azure resources and billing."
26
+
- content: "Which statement best describes a foundation model in Microsoft Foundry"
27
+
choices:
28
+
- content: "A small, task‑specific model that must be fine‑tuned before it can perform any useful function"
29
+
isCorrect: false
30
+
explanation: "This describes small or domain‑specific models, not foundation models."
31
+
- content: "A benchmarking tool used to compare different model families"
32
+
isCorrect: false
33
+
explanation: "Benchmarks and leaderboards compare models, but they are not foundation models themselves."
34
+
- content: "A large, pretrained model that provides general capabilities and can be used immediately or customized"
35
+
isCorrect: true
36
+
explanation: "Foundation models are large, pretrained models that offer broad language, reasoning, or multimodal abilities out of the box. They can be deployed immediately or fine‑tuned."
37
+
- content: "In the Foundry portal, what is the primary benefit of using the Model Playground before writing code?"
38
+
choices:
39
+
- content: "It lets you test prompts, compare models, and capture working settings that you can reuse in code."
40
+
isCorrect: true
41
+
explanation: " The Playground is described as the easiest way to interact with a deployed model and a bridge to code: you can try prompts, compare models, tune key parameters (temperature, max output tokens, system instructions), and then reuse those same values in your client application."
42
+
- content: "It deploys the model for you and removes the need to use an API."
43
+
isCorrect: false
44
+
explanation: "The Playground helps you interact with a deployed model and experiment with prompts and settings, but it doesn’t eliminate the need for an API in an application. When you move to a real app, you still call the deployed model through an endpoint (for example, via the OpenAI-compatible Responses API)."
45
+
- content: "It replaces system instructions by automatically generating the best system prompt for every scenario."
46
+
isCorrect: false
47
+
explanation: "System instructions (system prompts) are something you define to set behavior, tone, and guardrails for the assistant. The Playground allows you to configure and test system instructions, but it doesn’t replace them or guarantee an optimal one for every use case."
48
+
- content: "What is the primary outcome of publishing an agent in Microsoft Foundry?"
49
+
choices:
50
+
- content: "It converts the agent into a managed Azure resource with a stable endpoint that you can share and integrate without exposing your Foundry project or source code."
51
+
isCorrect: true
52
+
explanation: "Publishing promotes an agent from a saved development asset to a managed Azure resource with a stable endpoint. "
53
+
- content: "It automatically reduces costs by making the agent free to run, regardless of model tokens, tools, or connected data services."
54
+
isCorrect: false
55
+
explanation: "While the agent itself is described as “free,” real costs come from what it uses—model tokens, tools invoked, storage, connected data services, and deployment configuration. Publishing does not eliminate those costs."
56
+
- content: "It replaces the need for the Project API by allowing the agent to be called only through the Foundry portal UI."
57
+
isCorrect: false
58
+
explanation: "After publishing, you can integrate the agent anywhere by calling it from a client application using the Foundry Projects SDK and the Project API. Publishing expands integration options rather than limiting them to the portal."
59
+
- content: "In the Python example, which line is responsible for calling the published agent (rather than calling a model deployment directly) when generating a response?"
explanation: "This line only creates the client that can send requests. It sets up the capability to call the Responses API, but does not invoke the agent by itself."
67
+
- content: "`response = openai_client.responses.create(input=[{'role': 'user', 'content': 'Tell me what you can help with.''}], extra_body={'agent': {'name': agent.name, 'type': 'agent_reference'}},)`"
68
+
isCorrect: true
69
+
explanation: "This is the actual request that produces output. The key detail is the extra_body field with 'type': 'agent_reference', which tells the API to route the request to the agent (by referencing it), rather than treating it as a normal model-only call."
In a short few years, generative AI, a subset of artificial intelligence that focuses on creating new content, has changed the way we work and revolutionized what is possible with technology. At times, the fast-moving developments in generative AI can feel challenging to keep track of even for seasoned developers.
2
+
3
+
In this module, gain a framework for understanding generative AI applications and how Microsoft Foundry supports innovation. What does today's innovation look like? Consider these use cases:
4
+
5
+
-**Marketing Content Creation**: Companies use Microsoft Copilot's generative AI to automatically write product descriptions, blog posts, and social media content—saving time and ensuring brand consistency across platforms.
6
+
7
+
-**Customer Support**: Businesses deploy AI-powered virtual agents that can understand and respond to customer inquiries in natural language, offering 24/7 support and reducing the load on human agents.
8
+
9
+
-**Code Generation**: Developers use tools like GitHub Copilot to generate code snippets, suggest functions, and even write entire modules based on natural language prompts, speeding up software development.
10
+
11
+
-**Image and Video Generation**: Designers and content creators use the latest models in Microsoft Foundry's model catalog to generate visuals for campaigns, storyboards, or concept art—often from just a text description.
12
+
13
+
-**Personalized Learning and Tutoring**: Educational platforms use generative AI to create custom quizzes, explanations, and study guides tailored to a student’s learning style and progress.
14
+
15
+
Microsoft offers an ecosystem of tools for AI use and development. This module explores the Foundry model catalog and how to discover, evaluate, and deploy an appropriate model. Learn how to test and configure the deployed model the Foundry playground, and call it from code using the OpenAI‑compatible Responses API. Finally, you’ll see how agents encapsulate a model, its instructions, and optional tools so your solution is reusable and consistent across Playground and code via the Project API.
0 commit comments