M365 Copilot App Model Selector Active Inside Published Copilot Studio Agent — By Design or UI Gap?

(2) Share

Report

Posted on by MM-13041652-0

Hi everyone,

I published a Copilot Studio agent to the M365 Copilot application. Within Copilot Studio, I have a specific AI model selected for the agent's generative orchestration (e.g., Claude Sonnet 4.6). However, when I open the M365 Copilot app and click into my agent, the platform-level model selector (Auto / Quick Response / Think Deeper / specific models) remains active and allows me to choose a different model — even though I'm now inside the context of my Copilot Studio agent.

Is this by design? Does the M365 Copilot model selector intentionally apply to Copilot Studio agents published to M365 Copilot, or is it a UI element that should be hidden/disabled when a user enters an agent context?
If the selector is active, does it actually override the model I configured in Copilot Studio, or does the agent's model setting take precedence regardless of what the user selects?
If the selector does override the agent model, is there any way for makers to lock the model so end users cannot change it when interacting with the agent in M365 Copilot?

Has anyone else encountered this? Would love to hear if others have tested whether the model selector actually changes the agent's behavior or if it's purely a UI artifact.
Thanks in advance!

Categories:

Publish & channel management

I have the same question (0)

All responses (4)

Answers (0)

Sort by

Prasad-MSFT Microsoft Employee on at

Like
a
(1)

Report
Copy link

Link copied!

The M365 Copilot app’s model selector is intentionally available and will override your agent’s configured model. There is no way to lock the model at the agent level when published to M365 Copilot. This is a platform-level design choice.

Thanks
Prasad Das

Your feedback is important to us. Please rate us:

🤩 Excellent 🙂 Good 😐 Average 🙁 Needs Improvement 😠 Poor

Was this reply helpful? Yes No
Suggested answer

Haque 3,653 on at

Like
a
(1)

Report
Copy link

Link copied!
Hi @MM-13041652-0,

The behavior you’re seeing is expected based on how Microsoft 365 Copilot integrates with Copilot Studio agents:

When you publish a Copilot Studio agent to the M365 Copilot app, the agent’s orchestration and AI model selection inside Copilot Studio (e.g., Claude Sonnet 4.6) is fixed for that agent’s backend processing.

However, the M365 Copilot app itself includes a platform-level model selector that remains active and allows users to override or choose different models (Auto / Quick Response / Think Deeper / specific models) at runtime.

This model selector in the M365 Copilot app is a user-facing control that applies on top of the agent’s configured model, effectively allowing users to influence the model behavior dynamically.

The Copilot Studio agent’s configured model acts as the default or baseline, but the M365 Copilot app’s selector can override it for the session or query.

This design provides flexibility for end users to adjust model depth or style without republishing the agent, but it can cause confusion if you expect the agent’s model choice to be locked.

I am sure some clues I tried to give. If these clues help to resolve the issue brought you by here, please don't forget to check the box Does this answer your question? At the same time, I am pretty sure you have liked the response!

Was this reply helpful? Yes No
Suggested answer

11manish 3,333 on at

Like
a
(0)

Report
Copy link

Link copied!

Yes, this behavior is by design. The model selector you see in the Microsoft 365 Copilot interface is part of the Copilot shell UI and is not specific to Copilot Studio

agents.

It remains visible even when interacting with an agent. However, in most cases, it does not override the model configured in Copilot Studio—the agent continues

to use its defined orchestration model.

Currently, there is no way for makers to disable or lock the model selector in the UI, which can create confusion, but the agent’s configuration generally takes

precedence during execution.

Was this reply helpful? Yes No
Suggested answer

Valantis 6,735 on at

Like
a
(1)

Report
Copy link

Link copied!

Hi @MM-13041652-0,

The model selector in M365 Copilot is a platform-level shell UI control that remains active regardless of which agent you're in, and it does override the model configured in Copilot Studio.

To answer your specific questions:

1. Yes, by design. The selector is intentionally part of the M365 Copilot shell and isn't disabled when entering an agent context.
2. The M365 Copilot selector takes precedence over the model you configured in Copilot Studio.
3. There is currently no way for makers to lock or hide the model selector for end users when the agent is published to M365 Copilot. This is a platform-level control and not configurable at the agent level.

This is worth noting when choosing Claude Sonnet 4.6 or any specific model in Copilot Studio if consistent model behavior is critical for your agent users can change it at runtime and your agent will use whatever model they've selected instead.

If model consistency is a hard requirement, Teams is currently the more controlled channel since the M365 Copilot shell UI behaves differently there.

Best regards,

Valantis

✅ If this helped solve your issue, please Accept as Solution so others can find it quickly.

❤️ If it didn’t fully solve it but was still useful, please click “Yes” on “Was this reply helpful?” or leave a Like :).

🏷️ For follow-ups @Valantis.

📝 https://valantisond365.com/

💼 LinkedIn

▶️ YouTube

Was this reply helpful? Yes No