Microsoft is unveiling a new feature in its Visual Studio Code editor that leverages automatic AI model selection for enhanced performance. This automated model selection will allow the editor to choose between various models, including Claude Sonnet 4, GPT-5, and GPT-5 mini, for users of GitHub Copilot. While free users will benefit from this model flexibility, those on a paid plan will predominantly use Claude Sonnet 4.
This indicates a strategic preference by Microsoft for Anthropic’s AI models over the latest offerings from OpenAI, particularly GPT-5. Sources close to Microsoft’s development initiatives have revealed that the company has encouraged its developers to work with Claude Sonnet 4 for some time.
“According to our internal benchmarks, Claude Sonnet 4 is the model we recommend for GitHub Copilot,” stated Julia Liuson, the head of Microsoft’s developer division, in a June email to staff. Although this recommendation came prior to the launch of GPT-5, it appears that Microsoft’s stance has remained unchanged.
In addition to model selection, Microsoft is significantly investing in the development of its own AI technologies. “We’re embarking on substantial investments in our own infrastructure. Currently, MAI-1-preview has only been trained using a modest 15,000 H100s, which is relatively small in the context of our future plans,” commented Mustafa Suleyman, Microsoft AI’s chief, during a recent company town hall.
Moreover, reports suggest that Microsoft is planning to integrate Anthropic’s AI models into its Microsoft 365 applications. According to The Information, some features of the Microsoft 365 Copilot will leverage these models, particularly after finding that they surpassed OpenAI’s capabilities in programs like Excel and PowerPoint.
Last week, OpenAI and Microsoft announced a new agreement that could pave the way for the AI startup’s initial public offering. Since 2019, Microsoft has invested over $13 billion in OpenAI and has established a comprehensive revenue-sharing framework. The tech giant will now permit OpenAI to utilize competitor cloud services, with expectations of more details regarding the “next phase” of their partnership to be disclosed shortly.
