Cannot fine-tune a gpt-4o-mini based model on Sweden Central

Question

Cannot fine-tune a gpt-4o-mini based model on Sweden Central

ParisaTabassum-7447 20 Microsoft Employee

We already have one model deployed on our OpenAI resource in Sweden Central, which was fine-tuned based on the GPT-4o-mini model. We are attempting to deploy a new model with different data in the same manner. However, when we select the base model, the "Next" button is disabled. Interestingly, if we select the GPT-Turbo-35 model, the "Next" button works. According to the Azure documentation, fine-tuning based on the GPT-4o-mini model is available in Sweden Central, and we have successfully done this once before. Why can't we do it anymore? Is there a new policy change, or is there a restriction on the number of models based on the same model that can be deployed on a resource?

2 answers

Your answer

Answer 1

Prashanth Veeragoni 4,105 Microsoft External Staff

Hi ParisaTabassum-7447,

Azure OpenAI offers different deployment types: Standard, Global Standard (preview), and Provisioned Managed (preview). Each has specific regional availability and requirements.

Action Steps:

· Verify Deployment Type: Ensure that you're selecting the appropriate deployment type that supports GPT-4o-mini in Sweden Central.

· Consider Provisioned Managed Deployment: If Standard deployment is not available due to quota or other constraints, you might explore Provisioned Managed deployment, which requires specifying Processing Throughput Units (PTUs) and using the ProvisionedStandard SKU.

Please refer to: Can I deploy a fine-tuned GPT-4o in a different region after fine-tuning for more info.

Hope this helps. Do let me know if you have any further queries.

If this answers your query, do click Accept Answer and Yes for was this answer helpful.

Thank you!

ParisaTabassum-7447 20 Reputation points Microsoft Employee

2025-04-29T17:57:23.3333333+00:00

Hello @Prashanth Veeragoni ,

Thank you for your response. We have chosen the Quota Type "Standard" with the Pricing Tier Standard S0. The Quota Allocation column indicates 50K of 500K TPM. The other model deployed on this resource is using 50K TPM, so we should be able to use the remaining 450K TPM, correct?

For the other model that we trained and deployed, the Deployment Type is "Standard". However, we cannot start the fine-tuning process for the new model because the "Next" button is disabled. Therefore, we couldn't proceed to the deployment step.
Prashanth Veeragoni 4,105 Reputation points Microsoft External Staff

2025-04-29T18:14:42.8166667+00:00

Hi ParisaTabassum-7447,

Thank you for the update and additional clarification.

You're correct that with a Standard quota type and Pricing Tier S0, if your current deployment uses only 50K TPM out of 500K TPM, you should theoretically have enough capacity (450K TPM) to deploy and fine-tune an additional model.

However, since you're unable to proceed past the model selection step (i.e., the "Next" button is disabled when selecting GPT-4o-mini), here are some additional points and potential actions to help you move forward:

1.Model Availability or Regional Restriction

Even though Sweden Central supports gpt-4o-mini, not all model versions or SKUs may be consistently available for fine-tuning across all tenants or regions at any point in time due to internal service constraints or temporary limitations.

Action: Please verify that the version you're selecting (e.g., gpt-4o-mini-2024-07-18) is one of the supported versions listed in the Azure OpenAI fine-tuning documentation.

2.Portal UI Glitch or Cache Issue

There have been cases where the Azure Portal UI does not enable the "Next" button due to a temporary glitch or stale cache.

Action:

·       Try a different browser (Chrome/Edge/Firefox).

·       Open an incognito/private window.

·       Clear cache and cookies.

·       If the issue persists, try using the Azure CLI or REST API to initiate the fine-tuning job instead of the portal.

Hope this helps. Do let us know if you any further queries.

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Thank you!

Answer 2

ParisaTabassum-7447 20 Microsoft Employee

So I guess, the main question is, Can I deploy 2 gpt-4o-mini models in the same resource using Standard Quota Type?

ParisaTabassum-7447 20 Reputation points Microsoft Employee

2025-04-29T18:22:57.6966667+00:00

Ignore the last comment please. Saw your response later than this post. I will get back to you if I need further help
Prashanth Veeragoni 4,105 Reputation points Microsoft External Staff

2025-04-30T18:30:06.3066667+00:00

Hi ParisaTabassum-7447,

Following up to see if the above answer was helpful. If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let me know.

Thank You.
Prashanth Veeragoni 4,105 Reputation points Microsoft External Staff

2025-05-02T13:09:55.87+00:00

Hi ParisaTabassum-7447,

Just checking in to see if you have got a chance to see my response to your question in resolving the issue.

If you are still facing any further issues, please don't hesitate to reach out to us. We are happy to assist you.

Looking forward to your response and appreciate your time on this.

If you feel that your quires have been resolved, please accept the answer by clicking the "Upvote" and "Accept Answer" on the post.

Thank you!

Share via

Cannot fine-tune a gpt-4o-mini based model on Sweden Central

2 answers

Your answer