How Building with the Latest OpenAI Models is Advancing Copilot Studio Development

Big changes are coming to Copilot Studio and they’ll make a real difference for anyone building or managing agents day to day. 

Last month GPT-4.1 became the default model for all new Copilot agents, replacing GPT-4o. The result? Faster responses and better quality interactions right out of the box. Existing agents running GPT-4o will keep working until 26 November but you can switch over to GPT-4.1 now through the model selection settings. 

On top of that the GPT-5 family of models – GPT-5 Auto, GPT-5 Chat and GPT-5 Reasoning – are now available for use in deployed agents. These remain in public preview but they bring significant leaps in reasoning, adaptability and contextual understanding. It’s another step towards more capable, human-like Copilot experiences. 

These are the same OpenAI models that power ChatGPT, now available securely within Copilot Studio through the Microsoft ecosystem. That means you get the creativity and expressive power users associate with ChatGPT combined with the enterprise-grade governance, control and integration that Copilot delivers. 

So what does this mean for makers and developers? It’s all about choice and capability. You can now tailor agents more precisely to the task – whether that’s fast, consistent transactional support or complex open-ended dialogue that calls for deeper reasoning. For teams used to ChatGPT’s creative flair, this brings some of that same expressive strength directly into Copilot Studio. 

For those of us building Copilot solutions, it means more flexibility in how we design interactions and more freedom to experiment with model behaviour. We can now align Copilot performance with specific client or business needs – from concise operational bots to AI companions that can explore ideas and problem-solve in more natural ways. 

At MPowerUp we’re already exploring how these new models can elevate Copilot development by combining the creativity users associate with ChatGPT and the enterprise reliability that Copilot delivers. 

 If your team builds or manages Copilot agents now’s the time to test GPT-4.1 and the GPT-5 preview models. The next wave of AI capability is already here and understanding how these models perform in your environment will help you stay ahead of the curve.