OpenAI Confirms GPT-5.6 Is 'Preferred Model' for Microsoft Copilot 365

OpenAI confirmed on July 9 that its newly released GPT-5.6 model family will serve as the 'preferred model' for Microsoft Copilot 365. The company made the statement directly to TechCrunch, pushing back against growing speculation that Microsoft might reduce its reliance on OpenAI technology.

Recent reports have suggested Microsoft is evaluating multiple strategies, including developing its own large language models or onboarding additional third-party AI providers, to diversify away from its singular dependency on OpenAI. These reports have fueled widespread discussion about the future of one of tech's most consequential partnerships.

OpenAI's response was clear and definitive: GPT-5.6 will continue to power Microsoft's suite of workplace and productivity applications. This means hundreds of millions of Copilot users will run their daily workflows—from document generation to data analysis to meeting summarization—on GPT-5.6.

The statement carries significant weight for both companies. For OpenAI, locking in Microsoft as both its largest customer and investor guarantees immediate massive deployment scale and revenue for GPT-5.6. For Microsoft, while the company pursues AI diversification, continuing to use the strongest available models is critical for Copilot's user experience and retention.

Microsoft has invested over $13 billion cumulatively in OpenAI by 2025, making it the startup's most important strategic partner. The deep integration—from exclusive Azure cloud services to product-level embedding—means any separation would be costly and protracted.

Notably, even as Microsoft invests in its own models and other AI vendors, a complete near-term replacement of OpenAI remains unlikely. GPT-5.6's advantages in reasoning and generation quality are essential for Copilot's competitiveness.

The real question going forward is whether Microsoft will gradually shift non-core AI workloads to its own or third-party models while retaining OpenAI for flagship features—a hybrid approach that may be the most realistic outcome.