OpenAI names GPT-5.6 'preferred model' for Microsoft 365 Copilot amid breakup chatter

Microsoft made OpenAI's GPT-5.6 family the default across Microsoft 365 Copilot Chat, GitHub Copilot and Microsoft Foundry, with the three variants — Sol, Terra and Luna — auto-selected based on task fit. OpenAI publicly labeling GPT-5.6 the 'preferred model' for Copilot is being read as a signal that the OpenAI–Microsoft partnership is entering a new phase rather than unraveling, tamping down persistent 'divorce' chatter.

For end users, the upgrade promises better text handling and logical reasoning across Word, Excel, PowerPoint and Copilot Chat, with task-aware routing sending heavy reasoning to Sol and cost-sensitive work to Luna. Microsoft also rolled infrastructure updates in the same week — Site Recovery upgraded to handle 5x churn at 500 MB/second, and Chaos Studio entered preview.

The subtext is strategic hedging. Even as it deepens the GPT-5.6 integration, Microsoft is reportedly developing its own in-house MAI models to reduce reliance on third-party AI and cut costs — the same custom-everything impulse driving Meta's chips and DeepSeek's silicon. So the 'preferred model' framing cuts both ways: it reassures markets the OpenAI relationship is intact today while Microsoft quietly builds the capacity to route Copilot to its own models tomorrow. Given the week's dominant cost narrative — Altman himself stressing dollars-per-task — Copilot's automatic model selection is as much about margin management as capability. Watch how much Copilot traffic Microsoft eventually shifts to MAI once those models mature.