Customize your agent from a Chat Window?

Microsoft Developer presents a model comparison that changes only the deployment, keeping the evaluation the same so results can be assessed alongside cost and latency.

Comparing models with measured results

The episode uses the Microsoft Foundry model catalog to assess which model fits a particular use case. The comparison follows three steps:

• Swap only the model deployment.
• Run the same evaluation against both options.
• Capture cost and latency from real runs.

This approach makes model selection a measured decision rather than a preference for the biggest model. The comparison examines when a smaller model is suitable for production—and when it is not—with evaluation results, cost, and latency considered together.

Watch the full episode or visit the Microsoft Foundry portal.