Uploaded August 2026 | Updated September 2026, 2 weeks ago
AI agents rarely rely on a single model to solve every task.
In this episode of Model Mondays, we'll explore how model router helps developers apply a hill-climbing approach to agent development by continuously finding better models for different steps in a workflow.
Learn how to evaluate models, compare performance across providers and model families, and dynamically route requests to the model best suited for each task.
Through discussions and demos, we'll show how model router helps agents improve quality, efficiency, and cost by treating model selection as an ongoing optimization process rather than a one-time decision.
----------------------------------------------
- Explore the Resources - github.com/microsoft/model-mondays
- Read the Newsletter - https://aka.ms/model-mondays/newsletter
- Continue the conversation on the Discord - https://aka.ms/model-mondays/discord
0:00 - Introduction and Code of Conduct
0:59 - Welcome to Model Mondays
2:19 - What's new in Model Router
3:06 - What is Model Router?
7:41 - Reducing latency and improving response precision
11:57 - How prompt routing works
17:58 - Updated list of available models
20:45 - Automatic model failover
24:14 - Policy enforcement in the routing stack
25:44 - Benchmarking the router
27:54 - Multiple agents, one model deployment
30:11 - What is hill climbing?
32:27 - Hill climbing with the model router
36:19 - Workshop: Can routing save money?
54:49 - Model Releases: How to build my optimization
1:04:24 - Weekly highlights in Microsoft Foundry
1:09:32 - What's next week on Model Mondays
[eventID:27496]
AI agents rarely rely on a single model to solve every task.
In this episode of Model Mondays, we'll explore how model router helps developers apply a hill-climbing approach to agent development by continuously finding better models for different steps in a workflow.
Learn how to evaluate models, compare performance across providers and model families, and dynamically route requests to the model best suited for each task.
Through discussions and demos, we'll show how model router helps agents improve quality, efficiency, and cost by treating model selection as an ongoing optimization process rather than a one-time decision.
----------------------------------------------
- Explore the Resources - github.com/microsoft/model-mondays
- Read the Newsletter - https://aka.ms/model-mondays/newsletter
- Continue the conversation on the Discord - https://aka.ms/model-mondays/discord
0:00 - Introduction and Code of Conduct
0:59 - Welcome to Model Mondays
2:19 - What's new in Model Router
3:06 - What is Model Router?
7:41 - Reducing latency and improving response precision
11:57 - How prompt routing works
17:58 - Updated list of available models
20:45 - Automatic model failover
24:14 - Policy enforcement in the routing stack
25:44 - Benchmarking the router
27:54 - Multiple agents, one model deployment
30:11 - What is hill climbing?
32:27 - Hill climbing with the model router
36:19 - Workshop: Can routing save money?
54:49 - Model Releases: How to build my optimization
1:04:24 - Weekly highlights in Microsoft Foundry
1:09:32 - What's next week on Model Mondays
[eventID:27496]
