Microsoft is shaking up its AI playbook again, this time by inviting Anthropic’s Claude Sonnet 4 and Claude Opus 4.1 to join the party inside Microsoft 365 Copilot.
Until now, Microsoft’s workplace AI assistant has been almost exclusively powered by OpenAI’s models, but starting today, customers can tap Anthropic’s models for tasks in Researcher and Copilot Studio.
Think of it as Microsoft’s version of a buffet: OpenAI’s GPTs are still on the menu, but now you can try a Claude-flavored dish if you’re feeling adventurous.
“Copilot will continue to be powered by OpenAI’s latest models,” says Charles Lamanna, who runs Microsoft’s business Copilot team.
“But now customers can mix and match Anthropic models too.” If you’re building custom AI agents or running deep research, you can toggle between OpenAI’s reasoning chops and Claude’s own brand of brainpower with a single click.
In Researcher, a shiny new “Try Claude” button lets you swap to Claude Opus 4.1 on the fly, while Copilot Studio builders can orchestrate tasks that combine models from Anthropic, OpenAI, and others in Azure’s growing catalog.
The rollout starts today for Microsoft 365 Copilot customers in the Frontier program, with Copilot Studio users able to opt in as well.
And yes, it’s a little ironic that while Microsoft is cozying up to Anthropic, the Claude models themselves are still hosted on Amazon Web Services, aka Microsoft’s archrival in cloud computing.
For now, Microsoft is accessing Claude through the Anthropic API, but don’t be surprised if an Azure hosting deal appears down the road.
This move comes just a week after Microsoft quietly leaned on Anthropic for Visual Studio Code, where GitHub Copilot now often defaults to Claude Sonnet 4 for coding tasks.
Rumor has it Excel and PowerPoint are next in line, thanks to Claude reportedly outperforming OpenAI’s models in testing.
Microsoft’s message is clear: the more AI brains, the better. Your spreadsheets and slide decks might never know which genius is behind the curtain, and that’s exactly how Redmond wants it.
Is Microsoft’s decision to integrate Anthropic’s Claude models into Office 365 Copilot a smart hedge against over-reliance on OpenAI, or does it create unnecessary complexity for users? Should we be concerned that AI assistants are becoming black boxes where users don’t know which model is powering their work, or does this “best tool for the job” approach actually benefit productivity? Tell us below in the comments, or reach us via our Twitter or Facebook.
