GitHub Copilot's October 19 Model Retirements: Migration Map

October 19, 2026 is the scheduled retirement date for six models in GitHub Copilot. The change matters beyond the model picker: GitHub's September 18 notice says it covers Copilot Chat, inline edits, ask and agent modes, and code completions. Teams that depend on a specific model should check their settings and model policies now. The official notice establishes the scheduled date, but it does not by itself prove that every client has completed the removal at the moment this article is read.

The six models and GitHub's suggested replacements

These are GitHub's published alternatives, not a promise that the new model behaves identically or is available to every account. GitHub says the six retiring models are affected across its Copilot experiences. A developer using a named model in an editor, a team standardizing on a particular Copilot Chat model, or an administrator controlling permitted models should all review the change. This announcement concerns GitHub Copilot; it should not be generalized into a claim that each model is retired from every provider's own API.

Why Business and Enterprise administrators need to check policies

GitHub says the suggested alternatives are automatically enabled for Copilot Business and Enterprise customers under default model enablement, unless an administrator has disabled the global default or explicitly disabled the replacement. That distinction is critical: a published successor name is not proof that a managed organization can select it. GitHub's documentation also explains that model access can depend on the Copilot plan, client, and organization or enterprise restrictions. An enterprise-enforced setting can prevent an organization owner from changing availability locally.

A practical migration checklist for teams

1. Check whether any of the six retiring models are still selected in the Copilot experiences and editor clients your team uses.
2. Compare each affected selection against GitHub's October 19 migration map; do not substitute a model based on name similarity alone.
3. For Business or Enterprise, inspect the global default model policy and explicit model restrictions. Organization owners can review Copilot > Models when permitted by enterprise policy.
4. Confirm the proposed replacement appears in the model selector for the relevant plan and client before asking everyone to switch.
5. Run a short representative test of critical coding, review or agent workflows and document any behavior differences.
6. Tell teammates which replacements are approved and who to contact if their client does not show the expected model.

GitHub's organization instructions show how to navigate to the organization's settings, open Copilot, select Models, and enable an allowed model. If an enterprise owner has enforced a setting, an organization administrator may not be able to override it. Users on individual plans have different access controls, so the enterprise policy instructions should not be presented as universal. GitHub says no action is required simply to remove deprecated models after retirement; the useful work is checking and adopting supported replacements.

Do not confuse October 19 with the November 16 wave

GitHub also announced a separate November 16, 2026 retirement for Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra, with a stated exception for certain annual individual subscriptions to Opus 4.8. That is a different set of models and a different deadline. Teams tracking Copilot model dependencies should keep both dates in the migration calendar rather than assuming that an October update addresses November's changes. Our separate November migration guide covers that later wave.

What is confirmed, and what still needs checking

For developers, the immediate action is to stop relying on an outgoing Copilot selection and verify a supported option in the actual environment where work happens. For team administrators, the priority is policy visibility: a replacement that exists in GitHub's documentation may still be disabled for a particular organization. A short inventory and targeted workflow check can prevent avoidable disruption without turning a model retirement into a large migration project.