Support for GPT-5 / GPT-5.1 models in chat-with-your-data-solution-accelerator?

[ogis-ando-m]

Hello team,
Thank you for the great chat-with-your-data-solution-accelerator project.
I would like to ask about the current and planned support for GPT-5 series models (e.g., GPT-5, GPT-5.1) in this solution.

Is the accelerator currently compatible with GPT-5 or GPT-5.1 when deployed on Azure OpenAI Service?
If not, are there known limitations (e.g., dependency on Chat Completions API vs. Responses API)?

GPT-5.x models require newer Azure OpenAI API versions and are typically used via the Responses API.
Are there any architectural assumptions in this accelerator (such as Chat Completions–specific logic or parameters like max_tokens) that would prevent using GPT-5.x models?

This information would be very helpful for planning future upgrades and evaluations.
Thank you very much for your time and for maintaining this project.
Best regards,
Makoto Ando

[Roopan-Microsoft]

Hey @ogis-ando-m - Thanks for reaching out to us.

Currently, this accelerator is supported and maintained with gpt-4.1. We will check with the build team to see if there are any plans to extend support to gpt-5.x models and will keep you informed.

In the meantime, we would greatly appreciate it if you could try deploying with gpt-5.x models and let us know if you encounter any issues. We’ll be happy to assist with any challenges you face.

[ogis-ando-m]

Thank you for your quick response and for looking into this.

After deploying the GPT-5.1 model in Azure OpenAI, I updated the following parameter to the correct value in the App Service environment variables:

AZURE_OPENAI_MODEL = gpt-5.1
AZURE_OPENAI_MODEL_NAME = gpt-5.1
AZURE_OPENAI_MODEL_VERSION = 2025-11-13
AZURE_OPENAI_API_VERSION = 2024-10-21

After applying this change, when I enter a question in the web application’s GUI and submit it, the following error message is shown:

"An error occurred. Please try again. If the problem persists, please contact the site administrator."

In my opinion, this issue may be caused by the fact that when using GPT‑5 family models, the application (currently using the Chat Completions API) is not yet handling some parameter changes compared to GPT‑4 family models.
For example, the token limit parameter appears to have changed from:

max_tokens → max_completion_tokens

If the app is still sending max_tokens (or other GPT‑4.x-style parameters) when GPT‑5.x is selected, it could result in a request failure and the generic UI error.
Could you please continue investigating this and consider updating the accelerator so it can support GPT‑5.x models as well?

Best regards,
Makoto Ando

[Roopan-Microsoft]

Hey @ogis-ando-m - Thanks for the suggestion. We’ll raise this with our team and keep you posted once we receive their response.