How to deploy fine-tuned model in AI foundry to other subscriptions

Question

I have fine-tuned LLM model in AI foundry but I can only deploy it within subscription. How to deploy it cross-subscription? Could you please guide us or provide link if any documentation is available for this topic?

Answer 1

Hello Jayesh Sangle,

Welcome to Microsoft Q&A and Thank you for reaching out.

Today, Azure AI Foundry does not support deploying a fine-tuned model directly across subscriptions using the portal UI. Fine-tuned models can only be deployed within the same subscription where they were created when using the Foundry interface.

However, cross-subscription deployment is possible using supported programmatic methods. Azure provides a preview feature called the Copy Model API, which allows you to copy a fine-tuned model from one subscription to another (as long as both subscriptions belong to the same tenant). This is currently available only through the REST API or SDK, not through the portal.

Here is how the cross-subscription copy works:

In the target subscription, you generate a copy authorization token.

In the source subscription, you call the Copy Model API, passing that authorization token.

Once the model is copied, it will appear in the target subscription and can be deployed like any other custom or fine-tuned model.

If you cannot use this preview API, the fully supported alternative is to re-create the fine-tuning process in the target subscription. This means setting up a new AI Foundry workspace in the new subscription, uploading your dataset again, and running the fine-tuning job there. Once trained, the model can be deployed normally inside that subscription.

Also, please ensure permissions are correctly configured, since cross-subscription operations require RBAC roles in both subscriptions (e.g., Cognitive Services Contributor, AI Foundry Project Contributor). The identity generating the deployment or copy token must have rights in both the source and target subscriptions.

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

Thank you!