Microsoft Integrates Anthropic Models into 365 Copilot

Microsoft announced on Wednesday the expansion of its Microsoft 365 Copilot by incorporating Anthropic's Claude Sonnet 4 and Claude Opus 4.1 models, alongside existing OpenAI options. Organizations can enable access through the Microsoft 365 admin center, allowing customers to choose between OpenAI and Anthropic models when using Copilot’s Researcher agent or building agents in Microsoft Copilot Studio.

In Copilot Studio, users can build, orchestrate, and manage agents using Anthropic models for reasoning, workflow automation, and other agentic tasks. A drop-down menu facilitates switching between models from Anthropic, OpenAI, or the Azure Model Catalog without leaving the platform. Claude in Researcher is being rolled out through Microsoft’s Frontier Program to Copilot-licensed customers who opt in, with hosting outside Microsoft-managed environments subject to Anthropic’s terms and conditions.

The addition of Anthropic models marks Microsoft's effort to reduce dependency on OpenAI. Alongside integrating models from more providers, Microsoft has begun testing its in-house model, MAI-1-preview. Meanwhile, OpenAI announced a structural shift to a Public Benefit Corporation (PBC) with a valuation expected to exceed $100 billion, following a Memorandum of Understanding with Microsoft.

Microsoft has invested over $13 billion in OpenAI since 2019, initially securing exclusive rights to power OpenAI’s tools through Azure. However, these restrictions have relaxed, with OpenAI expanding its infrastructure and cloud partnerships to include Oracle, Google, and others. OpenAI plans to invest $300 billion in projects with Oracle, while NVIDIA announced a $100 billion investment in OpenAI for new data centers.