Microsoft Prioritizes Anthropic’s Claude Model Over OpenAI’s GPT‑5 in VS Code and Microsoft 365
Key Points
- Microsoft adds auto‑model selection to VS Code, favoring Anthropic’s Claude Sonnet 4 for paid Copilot users.
- Internal benchmarks cited by Julia Liuson recommend Claude Sonnet 4 over OpenAI’s GPT‑5.
- Mustafa Suleyman announces training on a 15,000‑GPU H100 cluster for Microsoft‑owned AI models.
- Anthropic models will partly power the‑next Microsoft 365 Copilot, outperforming OpenAI in Excel and PowerPoint.
- New Microsoft‑OpenAI deal allows OpenAI to use rival cloud providers and may facilitate an IPO.
- Microsoft’s total investment in OpenAI exceeds $13 billion since 2019.
Microsoft is rolling out an automatic AI model‑selection feature for Visual Studio Code that favors Anthropic’s Claude Sonnet 4 over OpenAI’s GPT‑5, especially for paid GitHub Copilot users. Internal guidance from developer division head Julia Liuson cites benchmark results recommending Claude Sonnet 4. At the same time, the company is investing heavily in its own AI training infrastructure and planning to integrate Anthropic models into Microsoft 365 Copilot, while a new partnership deal with OpenAI eases cloud‑provider restrictions and hints at future financial arrangements.
AI Model Selection in Visual Studio Code
Microsoft announced a new automatic model‑selection capability for its Visual Studio Code editor. The feature evaluates available large‑language models and picks the one judged to deliver the best performance for each coding task. For GitHub Copilot free users, the system can choose among Claude Sonnet 4, GPT‑5, GPT‑5 mini, and other models. Paid Copilot subscribers, however, will primarily receive Claude Sonnet 4. The guidance reflects a shift toward Anthropic’s models, a move described by internal sources as a “tacit admission” that Microsoft prefers Anthropic’s offerings over OpenAI’s latest GPT‑5 for development workloads.
Julia Liuson, head of Microsoft’s developer division, communicated internally that “based on internal benchmarks, Claude Sonnet 4 is our recommended model for GitHub Copilot.” Her email, dated June, preceded the public release of GPT‑5 but the recommendation has remained unchanged, underscoring the company’s confidence in Anthropic’s performance in code‑generation scenarios.
Investments in Microsoft‑Owned AI and Expansion into Microsoft 365
Microsoft’s AI chief Mustafa Suleyman highlighted the firm’s “significant investments” in its own AI training clusters during a recent employee‑only town hall. He noted that the preview model MAI‑1 was trained on a cluster of 15,000 H100 GPUs, a relatively modest scale in the broader AI landscape but indicative of Microsoft’s commitment to building proprietary capabilities.
Beyond the developer tools, Microsoft is preparing to embed Anthropic models into select features of its Microsoft 365 suite. Sources report that the upcoming Microsoft 365 Copilot will be “partly powered by Anthropic models,” after internal testing showed superior performance in Excel and PowerPoint compared with OpenAI alternatives.
New Deal with OpenAI and Ongoing Collaboration
In parallel with the shift toward Anthropic, Microsoft and OpenAI announced a new agreement that could pave the way for the startup’s initial public offering. Since 2019, Microsoft has invested more than $13 billion in OpenAI and maintains a complex revenue‑sharing arrangement. The latest deal grants OpenAI the flexibility to use rival cloud providers, a notable change from earlier exclusivity expectations. While the precise terms of the “next phase” of the partnership remain undisclosed, the arrangement signals continued collaboration even as Microsoft diversifies its model portfolio.