Microsoft.CognitiveServices/accounts/deployments: Preflight validation errors during deployment — Issue with API version 2025-06-01

Hi Noman,

Based on the behavior you described, I would isolate this as a model-deployment preflight issue rather than an AI Foundry account/project issue.

The strongest clue is that:

• The AI Foundry account deploys successfully.
• The account + three projects deploy successfully.
• az deployment group validate succeeds.
• az deployment group what-if succeeds.
• The stack what-if fails only when the model deployment resources are included.

That suggests the Microsoft.CognitiveServices resource provider is rejecting something during the model-deployment preflight performed through the deployment-stack path.

I would first verify quota/capacity for BOTH model deployments in Sweden Central. Error 715-123420 is fairly opaque and should not be assumed to mean only "unusual activity." Microsoft documentation/examples show that Cognitive Services model deployment failures can return this error when sufficient model quota/capacity is not available.

You can check the available usage/quota with Azure CLI, for example:

az cognitiveservices usage list \

--name <account-name> \

--resource-group <resource-group> \

--location swedencentral \

-o table

I would also verify that the exact model name, model version, SKU, and requested capacity are currently supported in Sweden Central.

As a diagnostic step, try separating the model deployments from the deployment stack:

1. Deploy/manage the Foundry account and projects with the stack.
2. Create one model deployment separately using Azure CLI.
3. If that succeeds, add the second deployment.
4. Compare the successful deployment properties (model version, SKU, capacity, etc.) with the resources defined in the stack.

Example:

az cognitiveservices account deployment create \

--name <account-name> \

--resource-group <resource-group> \

--deployment-name <deployment-name> \

--model-name <model-name> \

--model-version <model-version> \

--model-format OpenAI \

--sku-name Standard \

--sku-capacity <capacity>

If the exact same model/SKU/capacity succeeds through the direct deployment API but fails only through az stack group what-if, that would substantially narrow this to the deployment-stack preflight path rather than the Cognitive Services resource itself.

Since you have already confirmed that regular deployment-group validate/what-if succeeds against the same template, I would not spend much time changing the parent Foundry account configuration until the model deployment quota/SKU/version and direct-deployment test have been ruled out.

If you can post the two model deployment blocks (with identifiers/secrets removed), particularly model name/version, SKU and capacity, that should make it possible to narrow this further.

Hope this helps,

Walker

It is preflight validation error. current model deployments

Service:
Azure AI Foundry / Azure OpenAI / Cognitive Services

Issue:
Model deployment blocked during ARM resource-provider
preflight validation.

Error code:
715-123420

Resource type:
Microsoft.CognitiveServices/accounts/deployments

The subscription has sufficient model quota.

The deployment is blocked during provider preflight validation
before deployment execution.