Microsoft Adds GPT-5.1 as Experimental Model Option in Copilot Studio
The AMW Read
Routine model-lineup update for an established enterprise agent platform, gated to experimental early-release access rather than a new capability or player.
Microsoft Adds GPT-5.1 as Experimental Model Option in Copilot Studio
Microsoft has added GPT-5.1 as an experimental model option inside Copilot Studio, its low-code platform for building enterprise AI agents and agentic workflows, rolling out alongside OpenAI's own release of the model. The option is currently limited to U.S. customers on early release cycle Power Platform environments. Microsoft describes GPT-5.1 as bringing improved adaptability in thinking time across both chat and reasoning modes, and explicitly frames the integration as pre-general-availability: intended for customers to evaluate performance against their own use cases and existing models, not for production deployment, while Microsoft completes its internal quality evaluation gates.
The move keeps Copilot Studio's model lineup current within days of an OpenAI release, reinforcing Microsoft's position as a primary enterprise distribution channel for OpenAI's newest models even as the platform also supports agents built on other providers. Framing GPT-5.1 as an opt-in experimental toggle rather than a default lets Microsoft advertise day-one access to a frontier model while shifting evaluation risk onto customers running it in non-production settings, a pattern consistent with how the platform has rolled out prior OpenAI model updates.
For enterprise teams already building agents on Copilot Studio, the practical step is benchmarking GPT-5.1's reasoning and latency characteristics against currently deployed models in the early release environment before any migration plan, since Microsoft is explicit that production use should wait for general availability. For competing low-code agent platforms, the episode underscores that near-simultaneous access to OpenAI's latest releases remains a distribution edge tied directly to the Microsoft-OpenAI partnership rather than something rivals can easily replicate.
