Microsoft is expanding model choice in Copilot Studio, giving users the flexibility to select between OpenAI and Anthropic models when building and orchestrating agents. This empowers makers to experiment, test, and optimize workflows by matching the right model to the right task. Anthropic’s Claude Sonnet 4 and Claude Opus 4.1 are now rolling out in early release and will be available for production use by year-end, alongside OpenAI models.
With this update, Copilot Studio supports multi-agent orchestration, allowing multiple agents with different primary models to work together seamlessly. Users can also leverage model-specific strengths for different tasks such as OpenAI for FAQs and Anthropic for compliance checks while tailoring agent behavior through the prompt builder. Admins maintain full control, enabling or restricting Anthropic models via Microsoft 365 Admin Center and Power Platform Admin Center. If Anthropic models are disabled, agents automatically fall back to OpenAI GPT-4o, ensuring continuity.
Copilot Studio continues to default to OpenAI for new agents but now offers greater choice, flexibility, and control for advanced reasoning, automation, and deep contextual workflows. By empowering organizations to experiment with model selection, Microsoft aims to transform business processes and help users shape the future of agentic AI.
Leave a comment