GitHub brings Copilot model management to REST API as six models retire

On October 19, 2026, GitHub made a REST API for managing Copilot model policies generally available at four levels: enterprise, enterprise team, enterprise application, and organization. That same day, the company retired six models in most Copilot scenarios. For administrators, this means two practical tasks: check which models are allowed by policy and make sure saved workflows use supported replacements.

The new API lets administrators retrieve and update policies using GET and PATCH requests. GitHub describes it as a way to centrally manage model availability; enterprise settings can delegate control to organizations, teams, or applications, depending on the selected mode. Details are available in the Copilot model policy API announcement.

Which models were replaced

The October 19 list names Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini, and Grok 4.5. GitHub listed Gemini 3.8 Flash, GPT-6.1 Sol, GPT-6.1 Sol, GPT-6 Luna, GPT-6 Luna, and Grok 4.7, respectively, as alternatives. The company noted that it had announced the retirements in advance on September 18 and that there is no need to remove deprecated models manually. The list and recommendations appear in the model retirement notice.

GitHub’s replacement list indicates a migration path, but whether a specific alternative is available depends on the organization’s Copilot policies. A Copilot Enterprise administrator may need to allow a model separately in the settings. Users can check the resulting availability in the Copilot Chat model picker in VS Code or on github.com.

What administrators should check

First, compare the list of retired models with what is actually allowed at each management level. Then review internal instructions, task templates, automation, and team conventions that refer to old models by name. After that, enable the required alternatives in policies and check that they appear in users’ Copilot interfaces.

The REST API is useful when manually updating settings across many organizations, teams, or applications makes oversight difficult. Before a large-scale migration, it is sensible to save the current policies, apply changes to one group, and verify the result by reading the configuration. This approach makes it possible to compare the expected model list with the actual access settings.

This change affects administration and model access in GitHub Copilot. It addresses a practical question: how to manage model availability programmatically after the catalog changes.