Microsoft Expands Foundry Model Router to 28 Regions, Updates Model Pool

Expanded Regional Availability

Microsoft's Foundry Model Router has significantly increased its regional presence. It expanded from operating in only East US 2 and Sweden Central to 28 regions for global standard deployments and 21 for data zone deployments. This expansion aims to support broader global access and deployment options for users.

Updated Model Pool

The router's model pool has been refreshed with new additions and removals. Anthropic Claude Opus 4.8 and the GPT-5.6 family were added to the selection. Concurrently, several models, including gpt-5-chat, gpt-5.2-chat, gpt-5.3-chat, and DeepSeek-V3.1, were removed as they reached their end of life.

Automatic Updates and Potential Impacts

Teams using the default configuration for the model router will automatically receive these changes without needing to redeploy their systems. While this ensures API stability, it introduces potential behavioral changes as new models are introduced and old ones removed. These changes can affect answer styles, tool-selection, structured-output reliability, latency, token usage, and failure patterns, even if the response schema remains the same.

Teams that have configured a specific subset of models are not automatically affected by the pool refresh. They are insulated from these changes by default and must explicitly add new models if desired. The router offers three modes: Balanced (default, optimizes cost and quality), Quality (for critical work), and Cost (for high-volume tasks).