Microsoft said Anthropic’s Claude models began rolling out in Copilot Studio on October 17, expanding the assistant’s multi-model options for business users, according to a company blog post.
The move gives Copilot Studio customers another large model provider for orchestrating agents and running chat and reasoning workloads, easing reliance on OpenAI infrastructure inside Microsoft’s flagship productivity assistant, the post said.
According to Microsoft’s Copilot blog, Anthropic models are now available in “early release cycle environments” worldwide and will reach preview in all environments within two weeks, with production readiness planned “by the end of the year.” Automatica reported in July that Microsoft had begun positioning Copilot as a multi-model system to avoid depending on a single AI vendor for core assistant capabilities.
Microsoft said Copilot Studio will keep OpenAI as the default model for new agents but will now allow customers to choose Anthropic’s Claude Sonnet 4 and Claude Opus 4.1 for orchestration and prompt-based chat scenarios. In the product’s prompt builder, users can select the “optimal model for each scenario” from a dropdown menu, the company said.
Microsoft described two integration paths. Customers can build, orchestrate and manage agents powered by Anthropic models for “advanced reasoning, workflow automation, flexible agentic tasks, and tool use.” They can also assign Anthropic models to specific prompts within Copilot Studio so different tasks use different providers, the post stated.
Administrators will control how Anthropic models are used. Anthropic must first be explicitly enabled in the Microsoft 365 Admin Center before it can be used across a tenant, after which it is on by default in the Power Platform Admin Center, where admins can further manage access for Copilot Studio makers, Microsoft said. If Anthropic is later disabled, agents built with those models will automatically fall back to the default OpenAI GPT-4o model, with “no additional configuration required,” according to the blog. The announcement does not include pricing details for Anthropic model use within Copilot Studio.
Microsoft framed the addition as part of a broader push for “model choice” in Copilot Studio and said customers can also use multi-agent orchestration to coordinate agents that each rely on different primary models.