If you’re new to AI, it’s easy to assume the biggest model in the catalog is the best choice for a chatbot or app. It usually isn’t. A bigger model can be slower, cost more, and still give inconsistent answers if it isn’t set up for the job.
The third video in my AI-901 Microsoft Azure AI Fundamentals series shows a better approach. It’s built around four questions you can ask up front: what’s the job, which model can do it, where and how does it run, and how should it behave? Answer those, and you end up with a deployment that’s faster, more accurate, and cheaper.
Start With the Job
The video opens with the six AI workloads you’ll see on the AI-901 exam: generative AI, agentic AI, text analysis, speech, computer vision, and information extraction. Each one comes with a practical example and where it lives in Azure, from Foundry Models and Foundry Agent Service to Azure Language, Speech, Vision, and Content Understanding in Foundry Tools.
It also tackles the workloads that are easy to confuse. What’s the difference between text analysis and a generative summary? When does computer vision end and information extraction begin? And if one big multimodal model can do it all, why use a specialized service? These are exactly the distinctions exam scenarios are built on.
Match the Model to the Capability
Next, the video walks through the model families you need to recognize, including large language models, reasoning models, small language models, multimodal models, and embedding models. You’ll learn two quick questions that narrow the choice and why starting with the smallest model that does the job well usually beats reaching for the newest one.
Compare Models in Microsoft Foundry
With more than 10,000 models in the Microsoft Foundry model catalog, you need a way to build a shortlist. The video shows how to use catalog filters and the model leaderboard to compare quality, safety, throughput, and cost, and how model router can pick the right model for each prompt automatically.
Choose a Deployment and Set the Parameters
Deployment types are one of the most testable topics in this part of the exam. The video breaks them down into two simple questions: where is the data processed, and how is it paid for? You’ll see how Global, Data Zone, and Regional options affect data residency, and when to choose Standard, Provisioned, or Batch.
Finally, it covers the configuration parameters that shape how a model behaves, including temperature, top P, max output tokens, and stop sequences, plus why reasoning models use reasoning effort instead of temperature.
Put It All Together
The video wraps up with real-world scenarios, from a call center that needs transcripts and sentiment scores by morning to a field technician app that has to work offline.
Watch the Video
If you’re preparing for the AI-901 exam or deciding which model to deploy in Microsoft Foundry, this video gives you a repeatable way to make the right call. Watch it now, and you’ll walk away with the four questions that make every model decision easier.
And if you find it helpful, subscribe to the channel and turn on notifications so you don’t miss the next video in the AI-901 series, where the focus turns to how AI reads text.
Links:
A Beginner’s Guide to the AZ-900
https://www.udemy.com/course/beginners-guide-az-900/?referralCode=C74C266B74E837F86969
Zero to Hero with Azure Virtual Desktop
Hybrid Identity with Windows AD and Azure AD
Windows 365 Enterprise and Intune Management
Exam AI-901: Microsoft Azure AI Fundamentals
https://learn.microsoft.com/en-us/credentials/certifications/exams/ai-901
Study guide for Exam AI-901: Microsoft Azure AI Fundamentals
https://learn.microsoft.com/en-us/credentials/certifications/resources/study-guides/ai-901
Explore Microsoft Foundry Models
https://learn.microsoft.com/en-us/azure/foundry/concepts/foundry-models-overview
Deployment types for Microsoft Foundry Models
https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/deployment-types
Model leaderboards in Microsoft Foundry portal
https://learn.microsoft.com/en-us/azure/foundry/concepts/model-benchmarks
