Capacity allocation request for gpt-image-2 model deployment

Question

Hello,

We are working to deploy the gpt-image-2 model for our visual production workflow following the retirement of DALL-E 3.

Currently, standard quota requests through the portal are showing as unavailable across regions due to current capacity limits. As our organization operates in the United States, our preference is to deploy in a US region (such as East US or Central US).

Could an Azure engineer please assist us with manual capacity allocation or review our request? We are happy to provide our subscription ID and organization details in a private message/comment upon request.

Thank you for your help.

Answer 1

Hello John,

The behavior you're seeing can occur when the requested model/region combination has no currently available deployment capacity for your subscription.

For gpt-image-2, Microsoft currently documents quota/capacity separately by deployment type and region. The Foundry portal and Model Capacities API can be used to determine where capacity is currently available for your subscription.

I would recommend:

Open Microsoft Foundry → Manage → Quota and check gpt-image-2 with Show all enabled.

Check both DataZone Standard and Global Standard deployment options where applicable.

Use the Model Capacities API to check availability across regions rather than repeatedly submitting deployments to the same region.

If quota is available but the deployment still cannot be created, submit a quota increase/support request with the subscription ID, target region, model/version, deployment type, and requested capacity.

Microsoft notes that quota increase requests are processed individually and that some model/region combinations may not support quota increases.

If the portal shows zero available capacity rather than simply insufficient quota, this is generally a capacity-availability issue rather than something that can be resolved by increasing the deployment's TPM allocation. In that situation, an Azure support request is appropriate for a subscription-specific capacity review.

For security, don't post the subscription ID or organization details publicly in the Q&A thread. Provide those details only through the official Azure support channel if requested.