GitHub Copilot is retiring 6 models on 19 October 2026: what to do

For more on working with models, see the Ailoot documentation.

Which GitHub Copilot models are being retired on October 19, 2026?#

The GitHub announcement dated 18 September lists the models and suggested alternatives. This is a migration list, not a promise that two models will give identical answers.

All rows have the same date: 19 October 2026. The announcement applies to all ways of working with GitHub Copilot, including Copilot Chat, inline edits, ask and agent modes, and code completion. It does not imply that models with the same names are being retired by their developers or in other services.

What automatic model enablement means#

For Copilot Business and Enterprise, the suggested alternatives are enabled automatically if default model enablement is in effect. Exceptions apply if an administrator has disabled the general rule or a specific model. Once access is enabled, the model appears in the Copilot Chat model picker in supported ways of working with GitHub Copilot.

This condition concerns access. The announcement does not promise to change a pinned model selection, project settings, or an integration. On the contrary, GitHub asks users to update workflows and integrations before the retirement date. So the statement “the new model will be enabled automatically” is no reason to skip the checks.

GitHub's separate statement that there is no need to remove the old models manually after retirement also does not mean “there is nothing to do before retirement.”

When working with AI tools, it is useful to distinguish access to a feature from checking it on your own task.

What to check before retirement#

1. Check which model is selected in Copilot. Compare its name with the first column. If you use a different model, this particular list is not a reason to change it.
2. Find the suggested alternative. In your work organization, check both the model picker and the administrator's permission. If general automatic enablement is disabled, the announcement suggests enabling access through the relevant model's policy in Copilot settings.
3. Check your settings and integrations. If a model being retired is explicitly specified somewhere, find the documented way to select a model in that particular tool. Do not guess by inserting a name from the table as an API identifier: the announcement does not provide a universal migration command.
4. Choose an available alternative and carry out a familiar task. For example, ask it to change a small section of code, then review the changes and run the project's checks. This lets you check the transition on your own task. We have not conducted a comparative test of the models.

A completed check means that the required alternative is available, selected in your tool, and suitable for the work task you tested. A name appearing in the menu confirms access alone. If the model is unavailable, first check the organization's policy. You cannot consider the integration migrated yet.

GitHub advises GitHub Enterprise customers with questions or difficulties to contact their account team.

What the announcement does not establish#

In this post, GitHub does not promise that a user relying on automatic model selection will notice nothing, nor that a particular CI process will necessarily break. Without details about the specific integration, both statements are too strong. The article contains neither a guarantee of uninterrupted operation nor a universal warning about API breakage.

Model availability and plan limits are also different questions. Limits are covered in a separate guide, what is free in GitHub Copilot. To evaluate results on tasks, see how to read AI coding benchmarks, and if you are choosing a model beyond this list, take a look at the comparison of GPT-6 Astra and Claude Fable 5.1.

Start with your selected model and access to its alternative. Before changing working integrations, check the result on a small task that you can evaluate yourself.

For further work, visit Ailoot.