One model deployment to route them all

https://aka.ms/foundry-portal https://aka.ms/InsideMicrosoftFoundryPlaylist Sparkle Cupcakes traffic is wildly mixed. Some requests are as simple as “what flavors do you have?” Others are a 120-person catering order. So why force it all through one model? You deploy Microsoft Foundry’s model router. Then you point the agent at it by changing a single line. The router sends simple orders to cheaper models and hard ones to frontier models in real time. You log which model served each request. Then you evaluate the router as a unit to confirm quality holds while cost drops. 0:00 - Meet the Microsoft Foundry Model Router 0:35 - Supported Models and Deployments 1:08 - Configure Routing Priorities 1:40 - Test the Cupcake Agent 2:48 - Route Complex Reasoning Tasks 3:18 - Test General Knowledge Questions 3:52 - Monitor Usage and Model Selection #AIAgents #Microsoft #Azure #foundry