Unable to complete official AI-103 labs because GPT-5.2 Global Standard quota is 0 TPM

Cezary Malecki 0 Reputation points
2026-09-18T21:43:08.6333333+00:00

I am currently working through Microsoft’s official AI-103 – Develop AI apps and agents on Azure training and I am unable to complete the hands-on labs because no interactive Azure OpenAI quota is available on my new Pay-As-You-Go subscription.

I successfully created a Microsoft Foundry resource and project in Sweden Central. The project and model catalog work correctly.

The AI-103 lab requires deployment of GPT-5.2 (version 2025-12-11) using Global Standard. However, the deployment fails with:

“Insufficient quota. The selected deployment type Global Standard with version 2025-12-11 cannot be deployed to your current project due to insufficient quota.”

In the quota management page, GPT-5.2 Global Standard shows 0/0 TPM.

I have already submitted quota requests for:

  • GPT-5.2 Global Standard
  • GPT-5.2 Data Zone Standard – EUR
  • GPT-5.2 Data Zone Standard – US

All three requests were rejected with the message that capacity is not available and that I should expand my regional choices or wait 30+ days.

I also tested other interactive Azure OpenAI models, including GPT-4.1 and GPT-5.4, and Standard deployments also report insufficient quota.

Interestingly, GPT-5.4 Global Batch does have 50M enqueued-token quota available on the same subscription. This indicates that Azure OpenAI access is enabled, but interactive Standard TPM quota is not available. Global Batch cannot replace the interactive deployment required by the AI-103 playground and API exercises.

My questions are:

  1. How is a learner with a new Pay-As-You-Go subscription expected to complete the official AI-103 hands-on labs when the required interactive quota is unavailable?
  2. Is there an official Microsoft sandbox or training subscription for AI-103 that provides the required model quota?
  3. Is there another supported deployment/model that Microsoft recommends specifically for completing these AI-103 exercises?
  4. Can a small interactive quota be enabled for educational use when the normal quota requests are rejected due to capacity?

This is for low-volume Microsoft Learn training only, not for production workloads.

I would appreciate guidance from the Microsoft Azure AI / Foundry team, as the standard quota request process has already been exhausted.

Azure OpenAI in Foundry Models
0 comments

1 answer

Sort by: Oldest
Most helpful Newest Oldest
  1. Vinodh247-1375 44,801 Reputation points Volunteer Moderator
    2026-09-19T01:12:41.0466667+00:00

    Based on the information you shared, this appears to be a quota allocation and capacity availability issue rather than a configuration problem with your AI Foundry project.

    Azure OpenAI quota is assigned per subscription and by model/deployment type. Having quota available for a Global Batch deployment does not automatically provide quota for Global Standard deployments because those quota pools are managed separately. Microsoft documents that quota is allocated independently for specific model and deployment combinations. [learn.microsoft.com], [learn.microsoft.com]

    The key detail in your screenshots and description is that GPT-5.2 Global Standard shows 0/0 TPM. When a deployment has no available TPM quota assigned, deployment creation will fail with the "Insufficient quota" message you are seeing. If quota increase requests for GPT-5.2 Global Standard and GPT-5.2 Data Zone Standard were rejected due to capacity constraints, there is no project-level or deployment-level setting that can override that restriction. [learn.microsoft.com], [learn.microsoft.com]

    Your observation regarding GPT-5.4 Global Batch is also expected. Batch deployments use a separate quota mechanism designed for asynchronous processing and do not provide the real-time inference capability required by AI-103 playground and application exercises. [learn.microsoft.com]

    One additional point worth noting is that the current Azure OpenAI quota model is capacity-driven. Even with a valid PAYG subscription, access to a specific interactive model deployment depends on quota availability for that model and deployment type. Therefore, the fact that the subscription is enabled for Azure OpenAI does not guarantee that every model can be deployed immediately. [learn.microsoft.com], [learn.microsoft.com]

    Based on the documented behaviour and the evidence provided, I do not see a user-side workaround that would enable GPT-5.2 Global Standard deployment while the assigned quota remains 0 TPM. The remaining open questions are really training-content questions:

    • Whether AI-103 currently provides a Microsoft-managed sandbox or training environment for these exercises.
    • Whether an alternative interactive model is officially supported for completing the lab when GPT-5.2 Global Standard quota is unavailable.
    • Whether the lab instructions are being updated to account for quota availability constraints on newer subscriptions.

    At this point, the limitation appears to be external to your Foundry project configuration and related to quota availability for the required deployment type rather than an implementation error in the lab steps.

    Help make this community better for everyone: if this answer resolved your issue, please accept it or leave an upvote. If not, share more details in a comment so we can continue the discussion and find the right solution.

    Was this answer helpful?

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.