OpenAI says GPT-5.6 is now the preferred model in Microsoft 365 Copilot, bringing the new model family into Word, Excel, PowerPoint, Copilot Chat, and Cowork.
That makes the GPT-5.6 launch more than a frontier-model announcement. It is also a distribution event inside the productivity suite where many enterprises already manage documents, spreadsheets, presentations, and cross-functional work.
OpenAI says Microsoft 365 Copilot will use GPT-5.6 to help users draft and edit documents, analyze data, build presentations, and coordinate work in Cowork. Microsoft will access the models directly through the OpenAI API.
Microsoft’s Foundry documentation lists the GPT-5.6 series as new Azure OpenAI models: gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna. The same page lists each with Responses API support, Chat Completions API support, structured outputs, text and image processing, functions and tools, parallel tool calling, computer use, and a 1,050,000-token context window.
Preferred does not mean invisible
The operational question for buyers is what “preferred” means inside each tenant.
OpenAI’s post names the Microsoft 365 Copilot surfaces, but it does not say that every tenant, region, workload, or administrative setting will route every request to the same GPT-5.6 variant. Microsoft Foundry separately labels the GPT-5.6 series as preview in Azure OpenAI.
That distinction matters because Microsoft 365 Copilot is not one product surface. A Word drafting session, an Excel analysis, a Copilot Chat answer, and a Cowork task can have different data access, latency, cost, audit, and reliability requirements.
For IT and AI governance teams, the near-term work is not only asking whether GPT-5.6 is available. It is checking where the model is used, how model updates are communicated, what data boundaries apply, and how performance changes are measured against internal workflows.
Microsoft gets the broad enterprise test
OpenAI’s own launch positioned GPT-5.6 as a family with stronger performance per dollar and more capability on demand. Microsoft 365 Copilot gives that claim a very different proving ground.
Model benchmarks and API demos test capability. Productivity-suite rollout tests whether the model can improve repetitive work that people actually do every day: rewriting, analysis, summarization, planning, and document production.
The challenge is that value will be uneven. Some teams may see direct gains in spreadsheet analysis or presentation drafting. Others may mostly notice output style changes, slower or faster responses, or different failure modes in familiar prompts.
That makes internal measurement more important than vendor positioning. Teams should compare a small set of recurring tasks before and after the model change: document first-draft quality, spreadsheet reasoning accuracy, handoff quality in Cowork, and the amount of human cleanup needed before work can be shared.
