How to access latest models on Azure AI Foundry, not all are available to assign to agent

Question

Hey Everyone,

I am using Azure AI Agent Service with on AI Foundry, and currently when trying to deploy GPT latest models such as 5.2 and 5.1. In the model deployment page in the second image the models are shown, but when I deploy gpt-5.2-chat I'm not able assign the deployment to agents. In the first image, these are the deployments that are available to assign to the agent directly. I'm not sure whether this is part of the subscription we have currently or we don't have enough quota, so I wanted to see what steps would we need to access the different models. Thank you in advance!

Answer 1

Hi Omar Elhanafy

GPT‑5.2 or GPT‑5.1 in the Model Deployment screen only means your subscription and region are entitled to deploy those models for inference. It does not mean those deployments are automatically usable by Azure AI Agent Service. The Agent Service has a separate compatibility and allow‑list layer.

Azure AI Agent Service supports only a limited, validated set of models:

Agents are not a thin wrapper over Chat Completions. They depend on planner loops, tool-calling stability, memory orchestration, schema enforcement, and retry logic. Only models that have been fully validated against this orchestration stack appear in the “assign to agent” list. That is why the first image shows fewer models than the second one.

At the moment, GPT‑5.2‑chat, GPT‑5.2, GPT‑5.1, and GPT‑5.x Codex variants can be deployed and used via the Responses API / Chat Completion APIs, but they are not enabled as agent backends. Because of this, the Agent UI intentionally hides those deployments during assignment.

Your deployment is succeeding, which already proves quota and subscription eligibility are not the issue. The lock icon you see next to GPT‑5.2 / GPT‑5.1 indicates feature gating, not missing quota. Even Enterprise, EA, or MCA subscriptions behave the same way today

Since the models appear in your deployment catalog and can be deployed successfully in that region, the region is already supported. The restriction exists only at the Agent Service layer, not at the Azure OpenAI service level.

New GPT major/minor versions are always exposed first via direct APIs (Responses / Chat Completions). Agent enablement follows later, once tool execution, safety, and long‑running orchestration behaviors are fully validated. Until that validation is complete, the model remains unavailable for agent assignment.

no tenant‑side steps to unlock GPT‑5.2 for agents today:

There is nothing you can configure, request, or redeploy to change this behavior. Once Microsoft enables GPT‑5.2 / 5.1 for Agent Service, your existing deployments will automatically become selectable no redeployment will be required.

(recommended options):

If you need full Agent Service capabilities, use a currently supported model such as GPT‑4.1, GPT‑4.1‑mini, GPT‑4.1‑nano, or GPT‑5 (base) where available. If you specifically need GPT‑5.2 reasoning quality, use it directly via the Responses API and implement your own tool/memory loop. A common production pattern is to combine both: Agent Service with GPT‑4.1 for orchestration, and GPT‑5.2 for specialised reasoning or transformation calls.

Note:GPT‑5.2 and GPT‑5.1 are visible and deployable in Azure AI Foundry, but they are not yet supported by Azure AI Agent Service, and this is by design, not a quota or subscription issue.

https://learn.microsoft.com/en-us/azure/ai-foundry/agents/concepts/model-region-support?view=foundry-classic&tabs=global-standard

https://learn.microsoft.com/en-us/azure/ai-foundry/concepts/foundry-models-overview?view=foundry-classic

https://learn.microsoft.com/en-us/azure/ai-foundry/agents/whats-new?view=foundry-classic

Do let me know if you have any further queries.

Thank you!