No quota available for o3-deep-research?

Hi Stephen Cattaneo

You're encountering a deployment issue with the o3-deep-research model in the West US region, where the Azure UI reports insufficient quota despite your approved access. This typically happens due to quota fragmentation across STUs/PTUs, regional capacity constraints, or model-specific quota visibility limitations. Even if your subscription has sufficient total quota, it may not be properly allocated to the specific model/version combination. The West US region is known for saturation issues, especially with newer models like o3-deep-research. Additionally, the quota request form not listing this model is a known limitation, as some models require internal allowlisting and are not exposed in the UI.

To resolve this, first verify your STU/PTU allocation in the Azure portal and consider reallocating quota from other deployments. If possible, switch to a less constrained region like East US or West US 2. If deployment in West US is essential, escalate the issue via https://aka.ms/GetHelp, including your subscription ID, deployment type, model version, and a ramp plan. You may also need to confirm whether your subscription is allowlisted for this model in the target region.

Reference links : Azure subscription and service limits, quotas, and constraints

Hope it helps!

Thank you

verify your STU/PTU allocation

How do I do this?

The reference link pertaining directly to lists the following limits, which I'm not near:

1. Maximum of 200 total Azure AI services resources per region.
2. Maximum of 100 resources per region

Model deployments have TPM quotas. These seem to be per region. I have no other deployments in the westus.

You may also need to confirm whether your subscription is allowlisted for this model in the target region.

o3-deep-research is only available in two regions (westus & norway)

Hi Stephen Cattaneo

Sorry for Delayed response!

To verify your STU/PTU allocation in Azure AI services, you'll need to check several key areas of your Azure configuration.

First, it's important to understand that Provisioned Throughput Units (PTU) represent your model processing capacity and are specifically allocated per region and subscription. These allocations define your maximum deployment capacity per region and are crucial for managing your Azure AI resources effectively.

The structure of your allocation follows a hierarchical pattern where quotas are scoped at the subscription level rather than tenant level. Each region within your subscription maintains its own dedicated quota pool, with limits applying per region, per subscription, and per model/deployment type. This means you'll need to verify your allocations separately for both West US and Norway regions, as o3-deep-research is only available in these two locations.

To perform the verification process, start by checking your allowlisting status, particularly important for o3-deep-research as it requires special approval. Then navigate to the Azure Portal and locate your Azure AI resource under the "Quotas + Limits" section to review your PTU allocation numbers for each region. Use the Azure AI Foundry deployment experience to check real-time capacity availability, keeping in mind that capacity can vary dynamically based on customer demand.

It's essential to be aware of several key limitations while verifying your allocation. Your total resources per region are capped at 200, with individual deployments limited to 100 resources per region. Additionally, TPM quotas apply separately to each region, requiring careful management across your deployment landscape. While having sufficient quota is important, it doesn't guarantee immediate capacity availability, especially in the West US region which is known for saturation issues with newer models.

If you encounter quota issues during verification, consider attempting deployment in the Norway region instead of West US, scale down your deployment temporarily, or try deploying at different times when capacity might be more readily available. If problems persist, contact Azure Support for assistance with your specific allocation requirements. Remember that even if you're well below the maximum limits of 200 total Azure AI services resources per region and 100 resources per region, regional capacity constraints and model-specific quota visibility limitations can still affect your deployment capabilities.

Reference links : Azure OpenAI in Azure AI Foundry Models quotas and limits , What is provisioned throughput?

Hope it helps!

Thank you

Hi Stephen Cattaneo

Do you get any chance to check above response

Thanks