GPT-5.6 becomes Microsoft 365’s preferred model
GPT-5.6 is becoming the preferred model for Microsoft 365 Copilot in Word, Excel, PowerPoint, Copilot Chat and Cowork. OpenAI and Microsoft expect it to produce more complete documents, analyses and presentations using fewer instructions and resources.
OpenAI announces that GPT-5.6 is becoming the preferred model for Microsoft 365 Copilot, the assistant integrated into Word, Excel, PowerPoint, Copilot Chat and Cowork. The change brings OpenAI’s new model family to the tools millions of people use to write, analyze data and collaborate.
Being the preferred model does not mean it is the only option available. It means Copilot will prioritize GPT-5.6 for handling Microsoft 365 customers’ tasks, with the goal of delivering more complete results using fewer instructions and fewer resources.
What changes in each application
The update is designed to reduce the back-and-forth that AI often requires. Instead of correcting a request several times or rebuilding the result from scratch, Copilot should produce a first version that is closer to what you need.
- Word: helps draft, edit and improve documents with fewer rounds of instructions.
- Excel: enables deeper analysis and helps you move faster from data to conclusions.
- PowerPoint: turns initial ideas into more polished and visually appealing presentations with fewer manual instructions.
- Cowork: supports complex tasks involving multiple areas and helps coordinate work to produce more complete results.
- Copilot Chat: uses the same model for conversations and general tasks within the Microsoft 365 environment.
In practice, this could help you ask Copilot to turn notes into a report, find patterns in a spreadsheet, prepare a presentation from a document or coordinate a task that requires information from several teams.
Microsoft will also access the model through the API
Microsoft will not only use OpenAI models integrated natively into its products. It will also access GPT-5.6 directly through the OpenAI API, the interface that allows an AI model to be incorporated into other services and applications.
According to OpenAI, GPT-5.6 is designed to deliver more useful work per token. Tokens are the units of text the model processes, and their use affects both the cost and speed of a task. The company also highlights better performance per dollar and the ability to use more capacity when work is particularly complex.
For users, the change is not about learning another application. It means the tools they already use will incorporate a model that can better understand lengthy requests, analyze information in more detail and produce more developed drafts.
The point to watch will be how these promises translate into everyday results: fewer corrections in documents, more reliable analysis in Excel and more useful coordination between teams. The integration brings OpenAI’s advanced models into a significant part of daily work, but their value will depend on the quality of the responses and human review before using the results.