Microsoft reportedly preparing its servers to host OpenAI’s GPT-5

If you have been using ChatGPT from its initial days, you probably know how far the AI chatbot has come since its launch just over two years ago. Well, it seems ChatGPT is going to get a lot better with OpenAI’s new models. In fact, a new report suggests that engineers at Microsoft are preparing for the launch of OpenAI’s upcoming GPT-4.5 and GPT-5 models.

Microsoft is reportedly preparing for the launch of OpenAI’s GPT-4.5 and GPT-5 models

The news comes via The Verge, which yesterday reported that Microsoft is preparing its server capacity for the GPT-4.5 and GPT-5 models. Last year, when OpenAI released the GPT-4o model, it reportedly took some teams at Microsoft by surprise. The reason was that the GPT-4o model undermined Microsoft’s paid-for AI services on Azure, which include speech and translation features. However, the firm appears to be better prepared for the launch of GPT-4.5 and GPT-5, expecting no further surprises this time around.

For the uninitiated, Sam Altman—OpenAI’s CEO—recently hinted that the GPT-4.5 model could be released in just a few weeks. The upcoming frontier model, codenamed Orion, is OpenAI’s last non-chain-of-thought model. OpenAI previously planned to release it late last year. But, for some reason, the company pushed the launch window to early 2025.

While OpenAI has teased the power of GPT-4.5, it is also preparing for the release of the GPT-5 model. A source familiar with the Redmond tech giant’s plan told the publication that it is expecting GPT-5 in late May. This potential release window is in line with Altman’s previous promise of launching GPT-5 in a few months.

GPT-5 will most likely integrate OpenAI’s new o3-mini model

A few past reports have indicated that OpenAI wouldn’t release another standalone o3 model anymore. Instead, the company will integrate the newly launched o3-mini into GPT-5. By merging the GPT and GPT-o LLMs, OpenAI wants users to have less confusion about which models to pick for a particular task or query.

Altman recently tweeted on X, noting that he “hates the model picker” and believes in the “magic of unified intelligence.” So, the reports about unifying both series models could indeed be true.