0 TPM quota for all full-size GPT-5, GPT-5.x and GPT-6 models in every region, and my quota requests were denied

Question

0 TPM quota for all full-size GPT-5, GPT-5.x and GPT-6 models in every region, and my quota requests were denied

Metin Birdal 0

My subscription (Azure Sponsorship, quota tier "Free Tier") has 0 TPM

interactive quota (Global Standard, Data Zone Standard and Standard) for

every full-size model newer than GPT-4.1, in every region. This includes

gpt-6-astra, gpt-5, gpt-5-pro, gpt-5-codex, gpt-5.1, gpt-5.1-chat,

gpt-5.1-codex and the rest of the gpt-5.x models. The only interactive

quota I have for newer models is gpt-5.4-mini and gpt-5-mini. The quota

page shows 0/0 TPM for all the others.

I submitted quota requests for gpt-6-astra (Global Standard) in UK South, EastUS 2,West US,

and then Sweden Central. All were denied because the region had no

capacity. The deploy dialog in Foundry also shows every region as

unavailable.

I use these models with Codex in VS Code for developing and testing a

game (code review, debugging, implementing features, running tests).

This is development use, not production. I can accept any region, and a

small amount such as 100K-200K TPM for any one of these models would be

enough.

Questions:

Is quota for full-size GPT-5, GPT-5.x or GPT-6 models available at all for Free Tier sponsored subscriptions?
What does my subscription need in order to qualify (a higher quota tier, a different subscription type, payment history)?
Which models and regions should I list in a new quota request to have a realistic chance?

I can share my subscription ID privately.

Deleted

This comment has been deleted due to a violation of our Code of Conduct. The comment was manually reported or identified through automated detection before action was taken. Please refer to our Code of Conduct for more information.
SRILAKSHMI C 19,735 Reputation points Microsoft External Staff Moderator

2026-09-29T15:36:11.91+00:00

Hello @Metin Birdal

Thank you for providing the detailed information.

We understand that your Azure Sponsorship subscription currently shows 0/0 TPM for several full-size GPT-5/GPT-5.x and GPT-6 models, and that your quota requests for gpt-6-astra were denied due to regional capacity.

Regarding your questions:

Free Tier availability: The current quota documentation does not provide a blanket confirmation that every full-size GPT-5/GPT-5.x or GPT-6 model is available for Free Tier subscriptions. If the quota page shows 0/0 TPM, it means that no deployable quota is currently allocated for that specific model/deployment type on your subscription.

Why the requests were denied: Quota availability depends on the model, deployment type, subscription, and available capacity. A request being denied with regional capacity unavailable indicates that the requested capacity could not be allocated in that region at the time of review. This is separate from your application or prompt usage.

Subscription requirements: There is no documented payment-history requirement that guarantees access to these models. Quota allocation and approval depend on the applicable quota tier, model availability, and capacity. Some models may also have separate access or eligibility requirements.

Which models/regions to request: Since your use case is development/testing and you are flexible on region, we recommend specifying the exact model(s) you need and requesting the minimum required TPM, such as 100K–200K TPM. You can also check other supported regions and deployment types where quota/capacity is available before submitting a request.

Please note that Support cannot guarantee approval for a particular model or region, as allocation is subject to the available capacity at the time of review.

For the latest quota information and request process, please refer to the Microsoft Foundry quota documentation:

Microsoft Foundry Models quotas and limits

Quota request form / request process (as referenced in provided docs): https://aka.ms/oai/stuquotarequest

Limited access for Azure Direct Models (gated models may require approval): https://learn.microsoft.com/azure/foundry/responsible-ai/openai/limited-access

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

Thank you!

1 answer

Your answer

Deleted

This comment has been deleted due to a violation of our Code of Conduct. The comment was manually reported or identified through automated detection before action was taken. Please refer to our Code of Conduct for more information.

Answer 1

Hi @Metin Birdal

Thanks for the detailed write-up, it makes this easy to diagnose. The pattern you're seeing (0/0 for every full-size model in every region, mini models fine) is a tier gate, not a capacity problem. The "no capacity" text on the denials is just the generic reason, which is why switching regions changed nothing.

To your questions:

On a Free Tier subscription, quota for full-size GPT-5, GPT-5.x and GPT-6 models isn't available. Free Tier gets the mini models only. Sponsorship credits keep a subscription in that tier, since there's no paid billing relationship behind it.
You need the subscription to move to Tier 1. That comes from a paid billing relationship with usage history, meaning a Pay-As-You-Go or MCA subscription with actual invoiced spend. Open Foundry > Quotas and look at the quota tier line. If it shows a Tier 1 upgrade date, that's your path. If the date has already passed and models still show 0/0, that's a backend provisioning issue and needs a support request or a moderator here to route it privately. Payment history matters, region choice doesn't.
Don't submit more requests from this subscription, they'll keep getting denied. Once you're on Tier 1, request Global Standard rather than regional Standard, since it draws from global capacity and sidesteps regional denials. And ask for the model you'll actually use with Codex (gpt-5.1-codex or gpt-5-codex) rather than the flagship gpt-6-astra, smaller asks clear faster.

Practical shortcut: create a Pay-As-You-Go subscription with a credit card, put a Foundry resource there, and run your Codex dev work from it. At 100 to 200K TPM for development, the monthly cost is small, and you keep the sponsorship for everything else. Meanwhile gpt-5.4-mini handles Codex tasks surprisingly well for code review and tests.

If this helped, please click Accept Answer so others with the same 0/0 puzzle can find it.

References: https://learn.microsoft.com/en-us/azure/ai-foundry/openai/quotas-limits https://learn.microsoft.com/en-us/answers/questions/5964500/azure-openai-tier-1-pending-while-gpt-5-5-and-gpt