Microsoft Foundry - Azure OpenAI: Previously Fine-tuned models disappeared

Question

Microsoft Foundry - Azure OpenAI: Previously Fine-tuned models disappeared

Fabio Pulvirenti 25 Microsoft Employee

All the fine-tuned models we trained in the past in our resource located in NorthCentralUS just disappeared. They were based on 4o-mini and 4.1-mini.

The models are still trainable from python, I launched another training in fact.

I happened in the past and they came back after a few days.

Marcin Policht 109.6K Reputation points MVP Volunteer Moderator

2026-02-16T12:09:04.12+00:00

As per https://learn.microsoft.com/en-us/answers/questions/5516467/azure-foundry-models-got-wiped-out-on-their-own

Inactivity on custom models

After you deploy a customized model, if at any time the deployment remains inactive for greater than fifteen (15) days, the deployment is deleted. The deployment of a customized model is inactive if the model was deployed more than fifteen (15) days ago and no completions or chat completions calls were made to it during a continuous 15-day period.

Reference - https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/fine-tuning?tabs=azure-openai&pivots=programming-language-studio#clean-up-your-deployments-custom-models-and-training-files

Model retirement

Models in Azure AI foundry has some life cycle policies on respective version.

Post retirement of models, you cannot use existing model deployment

https://learn.microsoft.com/en-us/azure/ai-foundry/concepts/model-lifecycle-retirement

If the above response helps answer your question, remember to "Accept Answer" so that others in the community facing similar issues can easily find the solution. Your contribution is highly appreciated.

hth

Marcin
Anonymous

2026-02-16T12:16:31.6533333+00:00

Hi Fabio Pulvirenti

please provide the details requested through private message

Answer accepted by question author

0 additional answers

Your answer

Marcin Policht 109.6K Reputation points MVP Volunteer Moderator

2026-02-16T12:09:04.12+00:00

As per https://learn.microsoft.com/en-us/answers/questions/5516467/azure-foundry-models-got-wiped-out-on-their-own

Inactivity on custom models

After you deploy a customized model, if at any time the deployment remains inactive for greater than fifteen (15) days, the deployment is deleted. The deployment of a customized model is inactive if the model was deployed more than fifteen (15) days ago and no completions or chat completions calls were made to it during a continuous 15-day period.

Reference - https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/fine-tuning?tabs=azure-openai&pivots=programming-language-studio#clean-up-your-deployments-custom-models-and-training-files

Model retirement

Models in Azure AI foundry has some life cycle policies on respective version.

Post retirement of models, you cannot use existing model deployment

https://learn.microsoft.com/en-us/azure/ai-foundry/concepts/model-lifecycle-retirement

If the above response helps answer your question, remember to "Accept Answer" so that others in the community facing similar issues can easily find the solution. Your contribution is highly appreciated.

hth

Marcin
Anonymous

2026-02-16T12:16:31.6533333+00:00

Hi Fabio Pulvirenti

please provide the details requested through private message

Answer 1

Hi Fabio Pulvirenti

We have escalated this issue to our Product Group (PG) team they reviewed the details and informed that earlier few days ago users could not view existing fine‑tuned models or create new fine‑tuning jobs through the UI. The impact affected more than 10% of users, spanning multiple regions globally, including UAE North, France Central, Italy North, East US 2, and others

The rollback is in progress,Now it should be mitigated for the customers.

The issue impacting Azure OpenAI fine‑tuning workflows through the Microsoft Foundry UI has been fully mitigated. The service is operating normally in production following a rollback to the last known good configuration. Customers should now be able to submit new fine‑tuning jobs and view existing fine‑tuned models without error

confirmed with the customer in teams and now 4o-mini and 4.1-mini models are back.

Thank you