Microsoft adds Anthropic’s Claude to Copilot, allows users to switch between AI models
Microsoft has integrated Anthropic’s Claude Sonnet 4 and Claude Opus 4.1 into Microsoft 365 Copilot’s Researcher and Copilot Studio, letting organisations switch models per task while OpenAI remains the default.
On 24 September 2025, Microsoft said it had integrated Anthropic’s Claude models into Microsoft 365 Copilot, allowing enterprise customers to choose Anthropic in addition to OpenAI for specific tasks.
The rollout began in Copilot’s Researcher agent and in Microsoft Copilot Studio for building custom agents. OpenAI models remained the default.
What changed, exactly?
Microsoft confirmed that Claude Sonnet 4 and Claude Opus 4.1 were being added as selectable options. In Researcher, users who opted in could switch between OpenAI and Anthropic for deep‑reasoning work; in Copilot Studio, builders could select Anthropic models to orchestrate agents and workflows.
The company said Researcher with Claude was rolling out through the Frontier Program to licenced Microsoft 365 Copilot customers, and that Copilot Studio support would reach preview across environments within two weeks, with production readiness targeted by the end of the year.
How it works and who can access it
Administrators need to enable Anthropic models in the Microsoft 365 admin centre before employees can use them. In Copilot Studio, a drop‑down let teams pick models per task, mixing Anthropic, OpenAI and others from the Azure model catalogue as needed.
Microsoft noted that Anthropic models were hosted outside Microsoft‑managed environments and subject to Anthropic’s terms. Separately, reports indicated the models were hosted on Amazon Web Services, underscoring a cross‑cloud approach.
The move has expanded Microsoft’s multi‑model strategy in productivity software, offering customers more control over accuracy, speed and cost trade‑offs. It also signalled continued diversification beyond a single AI supplier, even as OpenAI remained central to Copilot.
What's new for Copilot users?
After an admin enables the option, Copilot users can toggle Anthropic in the Researcher agent and choose Claude Sonnet 4 or Claude Opus 4.1 for tasks that required strong reasoning. OpenAI models continue to power Copilot by default, with Anthropic offered as an additional choice that could be switched on a task-by-task basis.
In Copilot Studio, creators are able to build and orchestrate agents using Anthropic, OpenAI and other models, selecting the most suitable option for each step of a workflow. The rollout began through the Frontier Program, with Researcher support launching first and Copilot Studio features moving gradually from early release to preview and production over the course of the year.