Microsoft partners with Anthropic to supercharge 365 apps
▼ Summary
– Microsoft is adding Anthropic’s Claude Sonnet 4 and Claude Opus 4.1 AI models to Microsoft 365 Copilot, expanding beyond OpenAI models.
– These new models are initially available in the Researcher tool and Copilot Studio, with a “Try Claude” button for easy switching in Researcher.
– In Copilot Studio, users can build and manage AI agents powered by Anthropic models for tasks like deep reasoning and workflow automation.
– The Claude models are rolling out today via an opt-in program and are currently hosted on Amazon Web Services, not Microsoft Azure.
– Microsoft has started favoring Anthropic models in other products and suggests more powerful experiences are coming to Microsoft 365 Copilot.
Microsoft is significantly expanding the AI capabilities within its Microsoft 365 Copilot by integrating Anthropic’s Claude Sonnet 4 and Claude Opus 4.1 models. This strategic partnership introduces a crucial element of choice for enterprise customers, who until now have relied primarily on OpenAI’s technology. The integration allows users to leverage these powerful new models directly within the Researcher tool and the Microsoft Copilot Studio platform for building custom AI agents.
According to Charles Lamanna, president of Microsoft’s business and industry Copilot team, this move underscores the company’s dedication to incorporating leading AI innovations. He states that while Copilot will continue utilizing OpenAI’s latest models, the addition of Anthropic’s technology provides customers with greater flexibility. The integration is specifically tuned for professional workloads and tailored to meet diverse business requirements.
A notable change is the introduction of a “Try Claude” button within the Microsoft 365 Copilot app for users of the Researcher agent. This feature enables a simple switch between OpenAI’s deep reasoning models and Anthropic’s Claude Opus 4.1. Lamanna confirms that once users opt-in, toggling between the different AI providers will be a seamless process.
The expansion also reaches Copilot Studio, Microsoft’s environment for creating and managing AI agents. Developers can now build agents powered by Claude Sonnet 4 and Claude Opus 4.1, which are well-suited for complex reasoning, workflow automation, and various agentic tasks. A key advantage is the ability to mix and match models from Anthropic, OpenAI, and other providers available in the Azure model catalog, assigning specific models to the tasks they perform best.
Availability for these new options begins today through the Frontier program. The rollout is for Microsoft 365 Copilot licensed customers who choose to opt-in, both for the Researcher feature and within Copilot Studio. In a notable technical detail, the Anthropic models will be hosted on Amazon Web Services (AWS), with Microsoft accessing them via an API, despite AWS being a direct cloud competitor. This arrangement is similar to how other developers integrate Claude, though Microsoft has previously moved to host other models, like xAI’s Grok 3, on its own Azure cloud, suggesting a future shift for Anthropic’s models is possible.
This announcement follows closely on the heels of Microsoft’s decision to favor Anthropic’s models within Visual Studio Code, where GitHub Copilot paid users now primarily use Claude Sonnet 4. Furthermore, industry reports indicate that Anthropic’s models, after demonstrating superior performance in internal tests, are expected to be integrated into core applications like Excel and PowerPoint in the near future. Lamanna hints that this is merely the start, promising that Anthropic’s technology will soon enable even more powerful experiences across the Microsoft 365 ecosystem.
(Source: The Verge)
