Microsoft is expanding the intelligence and flexibility behind its Copilot ecosystem by adding Anthropic models alongside its existing OpenAI-powered foundation.
Copilot will continue to be powered by OpenAI’s latest models, but Anthropic’s Claude Sonnet 4.5 and Claude Opus 4.1 are now available in Researcher or when building agents in Microsoft Copilot Studio.
By combining OpenAI’s broad set of models with Anthropic’s deep reasoning strengths, Microsoft is giving organizations more flexibility than ever to match the right AI model with the right job.
However, enabling these models isn’t as simple as flipping a switch. Claude is an externally facing model, so organizations risk inadvertently exposing sensitive data if they aren’t informed about its implications.
Copilot Studio Just Got Smarter (and Riskier): 7 LLMs Now Available
Today marks a significant step toward a multi-model future. Whether you need fast, general-purpose support or advanced reasoning for complex research, Microsoft now offers five OpenAI models and two new Anthropic models to choose from.
Let’s briefly overview all seven models, their best use cases, and what organizations need to know before enabling Claude.
OpenAI Models
Hosted within Microsoft’s secure cloud environment, backed by Microsoft’s compliance and data-handling protections.
1. GPT-4o (default)
A fast, general-purpose model that handles both text and images. It’s Microsoft’s default model, balancing accuracy, speed, and cost.
Best for:
- Customer service bots that summarize chats and calls
- Meeting recaps and quick drafting tasks
- Workflows that require text + screenshots or visual context
2. GPT-4.1
An upgraded version of GPT-4o with stronger reasoning and greater accuracy, focused on text-only tasks.
Best for:
- Reviewing policy and regulatory documentation
- Complex knowledge base Q&A
- Scenarios where accuracy is most vital
3. GPT-5 Chat
Built for conversation, this model provides natural dialogue and better context awareness, making interactions feel more human.
Best For:
- Employee self-service portals
- HR, IT, or benefits helpdesk agents
- Sales/marketing assistants that need to “chat like a person”
4. GPT-5 Auto
Optimized for carrying out multi-step workflows and orchestrating actions across systems, not just answering questions.
Best For:
- End-to-end workflow automation (e.g., create a ticket → assign → notify)
- Complex task routing across multiple systems
- Agents that act as “digital project managers”
5. GPT-5 Reasoning
Designed for complex problem-solving, this model can work through layered logic and multi-step analysis before delivering insights.
Best For:
- Root-cause analysis in operations or IT
- Risk modeling and forecasting in finance
- Supporting research-heavy roles like scientists and engineers
DOWNLOADABLE RESOURCE
Copilot Security Checklist: 25 Questions Every Leader Should Ask
Copilot is secure by design, but is your organization ready for it? Download our 25-question checklist to find out.
Anthropic Models
Hosted by Anthropic and not subject to Microsoft’s compliance policies.
6. Claude Sonnet 4.5
Anthropic’s most advanced model, designed to excel in coding, complex agentic workflows, and real-world computer use.
7. Claude Opus 4.1
Anthropic’s flagship reasoning model is optimized for deep reasoning, data analysis, and complex research.
Before your organization can enable Claude, an admin must explicitly review and accept Anthropic’s Terms of Service and data-handling policies in the Microsoft 365 Admin Center.
Important Note on Copilot Governance for Anthropic Models
While model choice brings flexibility, it also introduces considerations around data governance and security.
- OpenAI models are securely hosted within Microsoft’s cloud infrastructure, benefiting from Microsoft’s built-in compliance, security, and data-handling protections.
- Anthropic models are hosted outside Microsoft’s secure environment, so your organizational data is subject to Anthropic’s Terms of Service and not Microsoft compliance guarantees.
Because of this, Claude Sonnet 4.5 and Claude Opus 4.1 are positioned as experimental models. They’re available today through Microsoft’s Frontier Program, but we don’t recommend using them for customer-facing or regulated production use cases.
Production-grade Copilot use cases should continue to rely on Microsoft-hosted OpenAI GPT models.
Until stronger security and compliance guardrails are established, administrators should exercise caution when and if they enable Claude models. This situation underscores the importance of robust governance in safeguarding organizational data.
Future-Proof Your Business with Microsoft’s Multi-Model AI Solutions
This multi-model approach opens the door to more intelligent, customizable, and enterprise-ready agentic systems. But choosing the right model is only one part of the equation.
Successful AI adoption also requires strong data access governance, a clear adoption plan, and a future-proof business technology strategy that keeps people at the center of digital transformation.
As a Microsoft Solutions Partner specializing in Microsoft Copilot and agentic AI, we’re committed to guiding your organization through the complexities of Copilot implementation and adoption. From early experimentation to enterprise rollout, we help leaders reduce risk, accelerate ROI, and build AI strategies that deliver real business results.
Connect with us today to start shaping your Copilot competitive advantage.








