Check your replacement model policy before some Copilot models retire on November 16

Which models retire, when, and where?

The retirement date is November 16, 2026, and it covers all of Copilot. The October 15 announcement lists Copilot Chat, inline edits, ask and agent modes, and code completions. The models do not disappear from chat while staying available for completions.

The supported models documentation marks all three replacements as GA. The Auto model selection table also includes them for cloud agent, Chat, CLI, and the GitHub Copilot app, and lists configurable reasoning for each.

Who gets the replacement models automatically?

In Business and Enterprise organizations with default model enablement turned on, the suggested replacements are enabled automatically. If an administrator turned off the global default or disabled a replacement model, that automatic step does not apply. In that case, enable Claude Opus 5.5, GPT-6.1 Sol, and GPT-6 Luna yourself in the model policies under Copilot settings.

The supported models documentation also notes that available models may be limited by model policies. If you have been approving each new model one at a time, these replacements fall under the same approval. When enterprise and organization policies are managed separately, check that both levels allow them. Teams that deliver managed settings through several channels can compare timing in our post on Copilot managed policy delivery.

What changes for annual individual subscribers?

Individual subscribers on an annual Copilot plan can keep using Claude Opus 4.8 after the retirement date. That exception applies to this one model only, so Claude Opus 5 and GPT-5.6 Terra still go away on the same day.

Developers who use both a personal annual subscription and a company Business seat may see different model lists depending on the account. If your team docs say something like "run this with Opus 4.8," that instruction no longer fits the company account, so update it before November 16.

Where should teams start fixing things before November 16?

GitHub asks administrators to move workflows and integrations to supported models before the retirement date. Start by finding places where a model name is likely written down. The steps below are our suggestion based on the announcement, and the actual locations differ by repository.

1. In organization and enterprise Copilot settings, confirm that default model enablement is on and that all three replacement models are allowed.
2. Search repositories and internal docs for model names such as Opus 4.8, Opus 5, and GPT-5.6 Terra to find leftover references in automation settings, team wikis, and onboarding docs.
3. For work that used GPT-5.6 Terra, decide by task type whether it goes to GPT-6.1 Sol or GPT-6 Luna.
4. Rerun a few of your common requests on the new model and check that response format and length still fit the next step.
5. If you manage spending limits, check the replacement models' request multipliers in the billing docs and adjust budget alerts.

Should you also check API schedules outside Copilot?

Copilot model retirements and each provider's API model retirements run on separate schedules. If a service calls models from the same families directly through an API, check those dates separately. We cover the Claude API side in our post on the Claude Sonnet 4.5 retirement and the OpenAI side in our post on the OpenAI API deprecation schedule.

What is the short version?

• Claude Opus 4.8, Claude Opus 5, and GPT-5.6 Terra retire across Copilot on November 16.
• Organizations with default model enablement turned off must enable the replacements themselves.
• Only Claude Opus 4.8 stays available for annual individual subscribers.

The announcement says Enterprise customers can contact their account team with questions. We checked the official sources on 2026-10-18.