Microsoft is preparing server capacity for new generative AI models from OpenAI — GPT-4.5 and GPT-5, according to a source familiar with the company’s plans. OpenAI CEO Sam Altman recently stated that GPT-4.5 will launch in just a few weeks, and Microsoft expects to host the new model on its servers as early as next week. GPT-4.5, codenamed “Orion,” will be the last OpenAI model without the use of the “chain-of-thought” method. It is expected to be significantly more powerful than GPT-4.
Meanwhile, OpenAI is already planning to release the GPT-5 model, which will include major changes. Altman noted that it will be “a system that integrates many of our technologies.” It will also feature a new reasoning model, o3, which OpenAI announced during its holiday announcements in December. GPT-5 is expected to be released at the end of May, aligning with Altman’s promises to launch the next-generation model within a few months.
Microsoft is also working on improving Copilot, its AI-based tool, to make it more user-friendly. This includes simplifying the model selection process for tasks or queries. The company has already abandoned the “creative,” “balanced,” and “precise” options that previously helped control the model’s output, and has added a “Think Deeper” button that uses OpenAI’s o1 reasoning model.
Given these plans, Microsoft may make numerous announcements at its Build event in May. This will allow the company to showcase its latest technologies to developers building AI-based applications and products. Microsoft is expected to focus on enhancing the capabilities of its AI agents to make them even more attractive for business.
