Monday, November 16, 2026 Independent reporting on artificial intelligence Newsletter
AI News Beats
AI Tools

GitHub Copilot’s November model retirements reach their deadline

GitHub set November 16 for changes affecting three coding models, with a Claude Opus 4.8 exception for individual subscribers on annual plans.

A laptop screen displaying code against a dark blue backdrop
Photo: Nemuel Sereti / Pexels

GitHub Copilot users who rely on Claude Opus 5 or GPT-5.6 Terra face a November 16 retirement deadline for those models. GitHub has also scheduled Claude Opus 4.8 for retirement on the same date, with an exception for individual Copilot subscribers on annual plans.

The date comes from the GitHub Changelog, which says the changes span Copilot Chat, inline edits, ask and agent modes, and code completions. GitHub did not specify an hour for the change in the announcement. Its separate supported-models documentation still listed the affected names at the time of review, so the stated deadline should not be read as proof that removal has already reached every account.

Which replacements GitHub recommends

In its October notice, GitHub points users of Claude Opus 4.8 and Opus 5 toward Claude Opus 5.5. For GPT-5.6 Terra, it suggests either GPT-6.1 Sol or GPT-6 Luna. The recommendations are for use within Copilot; the notice does not establish that any of the underlying models is being removed from its original provider’s other products.

GitHub says the annual individual-plan exception preserves access to Opus 4.8 for those subscribers. It does not extend that exception to Opus 5 or GPT-5.6 Terra. That means two developers with otherwise similar Copilot workflows can see different model choices if they use different plans.

Organizations may need to check their policies

GitHub says suggested replacements are enabled automatically in Copilot Business and Enterprise under default model enablement. That does not apply where administrators have turned off the global default or expressly disabled a particular model. In those cases the administrator can check the organization’s model policies to see what it allows.

GitHub Docs also breaks out support by Copilot feature, client and subscription plan. A team that uses both a code editor and the Copilot command-line interface should check both rather than assuming that a model visible in one will appear in the other. An example is a team that handles chat in an editor but runs scripted jobs outside it: the visible chat selection does not prove the scripted job can select the same model.

A separate change is due in December

GitHub’s November 13 changelog notice lists a different batch of retirements for December 14. That is not a grace period for the models on the November list. Keeping those dates separate helps teams track which workflows need a substitute now and which can wait for the next planned update.

GitHub says users do not need to remove retired entries themselves. The work is instead to find places that refer to the outgoing names and verify an available replacement in the client or integration that uses it. A quick test on an ordinary work item can show whether an apparently available replacement is also usable for the task the team relies on.

Stay on top of AI

Get the stories that matter in artificial intelligence, straight to your inbox.

Free. One email with the stories that matter. Unsubscribe any time. See our privacy policy.