Model router in Microsoft Foundry is now generally
available. Model router is an AI orchestration layer dynamically selects the
optimal model for each prompt. It now supports a broader range of models
including: GPT-4 family, GPT-5 family, GPT-oss, Deepseek-v3.1,
Llama-4-maverick-instruct, Grok-4, and Grok-4-fast. New customization modes –
cost savings, quality, and balanced-- allow developers to fine -tune routing
logic to best suit their unique needs.
Additionally, the model router will be added to Foundry
Agent Service providing a single, intelligent endpoint that dynamically routes
requests to the most suitable model based on task complexity, performance, and
cost. This feature simplifies integration, optimizes outcomes, and delivers
enterprise resilience to enhance Foundry’s capabilities in multi-model
orchestration.
Learn more.