Future of Azure OpenAI Agents vs. Responses API (threads, GPT-5 support)

Question

We are currently using Azure OpenAI Assistants/Agents and are trying to understand the roadmap:

• According to OpenAI’s documentation, the Assistants API is being deprecated in favor of the new Responses API.

Within Azure AI Agent Service, we still see support for threads and agents, which are no longer natively part of the Responses API.

Our challenge: the threads functionality has been very useful for conversation management, but if Assistants are being phased out, we are unsure whether Azure AI Agents will continue to rely on that layer or if it will be fully re-architected.

We also understand that Assistants are not updated with the newest models (such as GPT-5).

Questions:

Will Azure AI Agents receive support for the latest models (e.g., GPT-5) in the near future?

For long-term stability, should we be migrating everything towards the Responses API and implement our own conversation state management, or will Agents remain a first-class, supported service with thread support?

Put differently: is Azure AI Agents a future-proof service on its own, or primarily a temporary wrapper on top of the soon-to-be deprecated Assistants API?

We’d greatly appreciate clarification so we can make an informed migration decision.We are currently using Azure OpenAI Assistants/Agents and are trying to understand the roadmap:

According to OpenAI’s documentation, the Assistants API is being deprecated in favor of the new Responses API.

Within Azure AI Agent Service, we still see support for threads and agents, which are no longer natively part of the Responses API.

Our challenge: the threads functionality has been very useful for conversation management, but if Assistants are being phased out, we are unsure whether Azure AI Agents will continue to rely on that layer or if it will be fully re-architected.

We also understand that Assistants are not updated with the newest models (such as GPT-5).

Questions:

Will Azure AI Agents receive support for the latest models (e.g., GPT-5) in the near future?

For long-term stability, should we be migrating everything towards the Responses API and implement our own conversation state management, or will Agents remain a first-class, supported service with thread support?

Put differently: is Azure AI Agents a future-proof service on its own, or primarily a temporary wrapper on top of the soon-to-be deprecated Assistants API?

We’d greatly appreciate clarification so we can make an informed migration decision.

Answer 1

Hello Maris !

Thank you for posting on Microsoft Learn.

GPT‑5 is generally available in Azure AI Foundry, including gpt‑5, gpt‑5‑mini, gpt‑5‑nano, and gpt‑5‑chat with full support for the Responses API and Chat Completions API https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models

However, GPT‑5 models (including GPT‑5‑chat) are currently not yet supported within Azure AI Agent Service, meaning you cannot deploy them as agents just yet https://learn.microsoft.com/en-us/answers/questions/5522229/how-to-create-agent-in-azure-ai-foundry-with-gpt-5

Microsoft has announced that GPT‑5 will be available soon in the Foundry Agent Service, including capabilities like browser automation and Model Context Protocol integration https://azure.microsoft.com/en-us/blog/gpt-5-in-azure-ai-foundry-the-future-of-ai-apps-and-agents-starts-here

GPT-5 is supported for API access in Foundry, but agent service support is forthcoming not immediately available.

The Responses API is the new unified, stateful API that offers many of the same capabilities as the older Chat Completions and Assistants APIs, combining them with advanced features like tool calling, computer‑use, file search, function calling, and chaining via response IDs https://azure.microsoft.com/en-us/blog/announcing-the-responses-api-and-computer-using-agent-in-azure-ai-foundry

The Responses API is currently in general availability in many regions, and provides the core foundation for agentic workflows

https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/responses

OpenAI plans to deprecate the Assistants API by mid‑2026, fully transitioning to the Responses API

https://www.theverge.com/news/627556/openai-ai-agents-responses-api-agents-sdk

There are hints that Microsoft is working to reconcile key assistant‑like and thread‑like features within the Responses API so those familiar patterns can be maintained though some features still lag behind https://community.openai.com/t/responses-api-code-interpreter/1153442