Copilot Studio

Unable to add the GPT-5 Series Model

(0) Share

Report

Posted on by ST-17080950-0

Categories:

Calling actions from Copilot Studio

All responses (4)

Answers (0)

Sort by

Suggested answer

Prasad-MSFT Microsoft Employee on at

Like
a
(1)

Report
Copy link

Link copied!

This looks more like a Copilot Studio platform limitation/bug than a developer configuration issue. The error indicates the connector is sending the legacy max_tokens parameter, which GPT-5.x doesn't accept.
We checked the current Microsoft docs, and GPT-5+ is currently documented as not supported for the Copilot Studio BYOM/custom prompt model integration, so we wouldn't recommend changing random model/deployment settings on your side.

We have reached out to engineering team regarding this issue. Once we have any update from them, we will inform you.
Thank you!
Suggested answer

ST-17080950-0 2 on at

Like
a
(0)

Report
Copy link

Link copied!

Thank you Prasad for your quick response.

Looking forward to your next update once you receive the engineering team regarding this issue.

Best regards,

Srinivas T
Prasad-MSFT Microsoft Employee on at

Like
a
(0)

Report
Copy link

Link copied!

Microsoft currently states that GPT-5 family models and later (GPT-5, GPT-5 mini, GPT-5 nano, GPT-5.x, etc.) are not supported for Bring Your Own Model in Prompts. Therefore, if you're unable to add a GPT-5 model from Azure AI Foundry in Copilot Studio, that behavior is expected and is likely due to this limitation rather than a configuration issue.

Ref: Bring your own model for your prompts - Microsoft Copilot Studio | Microsoft Learn
Suggested answer

BilalDev_01 72 on at

Like
a
(0)

Report
Copy link

Link copied!

Hi, I ran into the same issue. The error seems to be caused by a parameter mismatch rather than the model deployment itself.The connection is sending the max_tokens parameter, but this model does not support it and expects max_completion_tokens instead. Since the Azure AI Foundry/Copilot Studio connection appears to be adding max_tokens automatically, the model connection fails.

I would suggest checking whether the selected connection type and API version fully support the GPT-5 deployment. If there is an option to configure the token parameter manually, change max_tokens to max_completion_tokens.

If the parameter is not exposed in Copilot Studio, this may be a current limitation or compatibility issue with the connector. In that case, trying a supported API/model configuration or raising it with Microsoft support may be necessary.

Has anyone found a working configuration for GPT-5 in Copilot Studio with the Azure AI Foundry connection?