GPT-5.1 Now in Copilot Studio

GPT‑5.1 now becomes available in Microsoft Copilot Studio as an experimental model for U.S. early release Power Platform environments. The update gives users access to a next-generation AI model that handles both chat and reasoning with more adaptability. The rollout presents a chance to test cutting-edge capabilities in non-production scenarios.

What’s new with GPT-5.1 in Copilot Studio

With GPT-5.1, the model choices offered in Copilot Studio now include “GPT-5.1 Chat” and “GPT-5.1 Reasoning”. Users building agents can pick the model suited to their workflow. Because this comes under an experimental label, Microsoft recommends testing in non-mission-critical contexts while evaluation continues. The update brings improved thinking time flexibility and better reasoning in agent workflows.

How users benefit: speed, depth and flexibility

Users will see three key benefits of GPT-5.1 in Copilot Studio:

• Faster responses for chat tasks: For straightforward prompts, the system can opt for the “chat” variant and deliver quicker answers.
• Deeper reasoning for complex workflows: When an agent must plan, analyse or orchestrate multiple steps, the “reasoning” variant gives it time to gather context and validate its output.
• Experimentation before general availability: Early release means makers can evaluate the model’s behaviour, test custom prompts, and compare results without full production risk.

Real-world scenarios where it helps

For companies building custom agents, GPT-5.1 unlocks improved workflows in these areas:

• In the Power Platform, an agent can scan uploaded documents, extract key data, and suggest next-step actions with more accuracy.
• A service desk bot can switch between providing fast typical responses (via the chat mode) and undertaking deeper diagnostic logic when needed (via reasoning mode).
• Makers can trial the new model in non-user-facing contexts to refine prompts before deploying at scale.

What you should keep in mind

Because this is an experimental release, a few caveats apply:

• The availability is limited to U.S. customers in early-release Power Platform environments.
• It’s recommended for non-production scenarios while evaluation gates remain open.
• Organisations should plan pilot deployments, compare the model’s performance against existing models, and prepare for full-scale rollout when general availability arrives.

Looking ahead

As GPT-5.1 becomes more widely available, organisations using Copilot Studio will gain more refined control over how their agents think and respond. This dual-mode model choice—chat vs. reasoning—gives developers and makers more flexibility. By experimenting early, teams can prepare for the broader shift toward smarter, more adaptable AI agents.