Insiders have revealed that Microsoft engineers are currently preparing server capacity for OpenAI's upcoming GPT-4.5 and GPT-5 models. OpenAI CEO Sam Altman recently confirmed that GPT-4.5 will be released within weeks, with Microsoft expected to host the new AI model as early as next week.

GPT-4.5, internally codenamed Orion, represents OpenAI's next-generation cutting-edge model and is also the company's last non-chain-of-thought model. OpenAI has suggested that GPT-4.5 may outperform GPT-4, but the company is simultaneously looking ahead to the more transformative GPT-5 model.

According to sources, Microsoft anticipates the release of GPT-5 by the end of May, aligning with Altman's promise of a new generation of models arriving in a few months. However, this date could shift due to changes in release plans. Earlier reports indicated that OpenAI initially planned to launch GPT-4.5 at the end of 2024 but postponed it to early 2025.

GPT-5 is anticipated to be the more significant release among these two models, described by Altman as "a system integrating many of our technologies." The model will incorporate OpenAI's new o3 inference model, which was teased during the company's "12 Days of Christmas" announcements in December last year.

Although OpenAI released o3-mini last month, the company no longer plans to launch o3 as a standalone model, instead integrating it into the GPT-5 system. This aligns with OpenAI's goal of combining its large language models to eventually create a more powerful model, potentially referred to as artificial general intelligence (AGI).

The updated GPT-5 system will notably enhance ChatGPT's usability by unifying OpenAI's o-series and GPT-series models, reducing user confusion when selecting models for specific tasks or queries. In a recent post, Altman stated: "We hate model selectors as much as users do, and we aim to return to a magical unified intelligence."

If OpenAI releases GPT-5 on schedule by the end of May, it will coincide with Microsoft's Build developer conference, which begins on May 19 and overlaps with Google's I/O event. Last year, Altman made an appearance at the Microsoft Build conference shortly after OpenAI launched GPT-4o, a free and faster model available to all ChatGPT users.

OpenAI's release of GPT-4o last year caught some Microsoft teams off guard because it undercut Microsoft's paid AI services on Azure, including voice and translation features. By contrast, the launches of GPT-4.5 and GPT-5 should not come as such surprises, with Microsoft likely rolling out Copilot updates soon after OpenAI's latest product release. Last October, Microsoft undertook a major redesign of Copilot, just months after the release of GPT-4o.

Similar to OpenAI, Microsoft has been working to improve Copilot's model selector so users don't need to choose AI models while interacting with the assistant. Microsoft previously offered options like "Creative," "Balanced," and "Precise" to help control model outputs, but these were removed during last year's Copilot update. In January, Microsoft added a "Deep Reasoning" button, leveraging OpenAI's o1 inference model to answer queries.

Microsoft is also developing its own version of the Operator AI agent, which can perform actions online for users. The Operator interacts with graphical user interfaces (GUIs) on web pages, automating tasks for users automatically. This concept is similar to macros, which automate repetitive tasks, or autopilot systems in airplanes that operate without constant human attention.

Microsoft is expected to announce a series of AI-related updates over the coming months, particularly at the Build conference in May, where it will showcase its latest technology to developers building AI applications and products. Microsoft has recently focused heavily on enhancing the capabilities of its AI agents, with teams reportedly working to reduce costs to make them more appealing to enterprises.

This year, both Microsoft Build and Google I/O conferences will occur simultaneously, leading to expectations that the two companies will compete fiercely in their AI announcements, especially given the planned release of GPT-5 around the same time.