Reduce AI Costs with Adaptive Model Routing

Not every AI request needs your most capable, or most expensive language model. Some tasks require deeper reasoning, while simpler requests can be handled faster and more affordably by a smaller model. In this video, I demonstrate how to configure adaptive model routing with Postman Fabric Gateway. Fabric Gateway sits between your AI agent and your LLM services. Using decision model like Jev, it evaluates each request and dynamically selects the appropriate model and reasoning effort. You'll see how Fabric Gateway: - Receives a request configured to use GPT-6 Astra - Evaluates the complexity of the task using Jev - Routes a simple request to GPT-5.6 Luna instead - Selects a low reasoning effort for that request - Makes the routing decision visible through gateway traces I'll also walk through configuring the Adaptive Model Router plugin, connecting Jev as the decision service, and applying the routing policy to an OpenAI route. Adaptive model routing helps you balance model capability, response speed, and cost without manually selecting a model for every request or agent turn. If you found this helpful, let us know in the comments—and tell us what you'd like to see next. #AIAgents #LLM #ModelRouting #Postman #FabricGateway