gpt-5 azurerm_cognitive_deployment models require version specified while gpt-4o does not

[grizzdominant]

Is there an existing issue for this?

• I have searched the existing issues

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave comments along the lines of "+1", "me too" or "any updates", they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment and review the contribution guide to help.

Description

When we deploy a gpt-4o model with no version specified, it defaults to one and deploy is successful:

$ terraform state show 'module.DDMI_AzureAI_0_1.azurerm_cognitive_deployment.deployments["gpt-4o"]'
# module.DDMI_AzureAI_0_1.azurerm_cognitive_deployment.deployments["gpt-4o"]:
resource "azurerm_cognitive_deployment" "deployments" {
    cognitive_account_id       = "/subscriptions/*****/resourceGroups/test-rg/providers/Microsoft.CognitiveServices/accounts/test-rg-openai"
    dynamic_throttling_enabled = false
    id                         = "/subscriptions//*****/resourceGroups/test-rg/providers/Microsoft.CognitiveServices/accounts/test-rg-openai/deployments/test-rg-gpt-4o" 
    name                       = "test-rg-gpt-4o"
    rai_policy_name            = "Microsoft.DefaultV2"
    version_upgrade_option     = "OnceNewDefaultVersionAvailable"

    model {
        format  = "OpenAI"
        name    = "gpt-4o"
        version = "2024-11-20"
    }

    sku {
        capacity = 250
        family   = null
        name     = "DataZoneStandard"
        size     = null
        tier     = null
    }
}

When we change the model name to "gpt-5" and do not specify a version, we encounter this error:

│ Error: creating Deployment (Subscription: "****"
│ Resource Group Name: "test-rg"
│ Account Name: "test-rg-openai"
│ Deployment Name: "test-rg-gpt-5"): performing CreateOrUpdate: unexpected status 400 (400 Bad Request) with error: DeploymentModelNotSupported: The model 'Format:OpenAI,Name:gpt-5' of account deployment is not supported.

Then when adding a specified version, the deploy is successful:

# module.DDMI_AzureAI_0_1.azurerm_cognitive_deployment.deployments["gpt-5-2025-08-07"]:
resource "azurerm_cognitive_deployment" "deployments" {
    cognitive_account_id       = "/subscriptions/686e8e2b-ebce-4cba-b892-ada94eeb69fe/resourceGroups/azc-e2-nprd-hosting-ai-rg/providers/Microsoft.CognitiveServices/accounts/azc-e2-nprd-hosting-ai-rg-openai"
    dynamic_throttling_enabled = false
    id                         = "/subscriptions/686e8e2b-ebce-4cba-b892-ada94eeb69fe/resourceGroups/azc-e2-nprd-hosting-ai-rg/providers/Microsoft.CognitiveServices/accounts/azc-e2-nprd-hosting-ai-rg-openai/deployments/azc-e2-nprd-hosting-ai-rg-gpt-5"  
    name                       = "azc-e2-nprd-hosting-ai-rg-gpt-5"
    rai_policy_name            = "Microsoft.DefaultV2"
    version_upgrade_option     = "OnceNewDefaultVersionAvailable"

    model {
        format  = "OpenAI"
        name    = "gpt-5"
        version = "2025-08-07"
    }

    sku {
        capacity = 250
        family   = null
        name     = "DataZoneStandard"
        size     = null
        tier     = null
    }
}

Can someone confirm if this is expected behavior for the more recent models such as gpt-5?

Terraform Version

1.13.5

AzureRM Provider Version

4.67.0

Affected Resource(s)/Data Source(s)

azurerm_cognitive_deployment

Terraform Configuration Files

resource "azurerm_cognitive_account" "example" {
  name                = "example-ca"
  location            = azurerm_resource_group.example.location
  resource_group_name = azurerm_resource_group.example.name
  kind                = "OpenAI"
  sku_name            = "S0"
}

resource "azurerm_cognitive_deployment" "example" {
  name                 = "example-cd"
  cognitive_account_id = azurerm_cognitive_account.example.id

  model {
    format  = "OpenAI"
    name    = "= "gpt-5"
  }

  sku {
    name = "DataZoneStandard"
  }
}

Debug Output/Panic Output

│ Error: creating Deployment (Subscription: "****"
│ Resource Group Name: "test-rg"
│ Account Name: "test-rg-openai"
│ Deployment Name: "test-rg-gpt-5"): performing CreateOrUpdate: unexpected status 400 (400 Bad Request) with error: DeploymentModelNotSupported: The model 'Format:OpenAI,Name:gpt-5' of account deployment is not supported.

Expected Behaviour

No response

Actual Behaviour

No response

Steps to Reproduce

No response

Important Factoids

No response

References

No response

[mpenndev]

I encountered this problem for GPT-5.2 & GPT-5.4.

But GPT-5.1 worked as expected with not providing the version and it defaulting successfully.

[rcskosir]

This looks like Azure API behavior rather than a provider bug. In the provider, the model.version field is defined as optional, and when it's not specified, the provider omits it from the API request entirely,letting Azure handle defaults server-side.

For established models like gpt-4o, Azure has a default version it can fall back to, so the deployment succeeds without one. For newer models like gpt-5, Azure likely doesn't have a default version configured yet, which is why the API returns DeploymentModelNotSupported when no version is provided.

So yes — specifying version explicitly for gpt-5 (and likely other newer models) is the expected workaround. You might consider always specifying version in your configurations to avoid this inconsistency across models.