All playbooks
Build GuideGitHub CopilotAI CodingModel Migration

GitHub Copilot Model Retirements: Fix Team Defaults

Check GitHub Copilot’s October 19 model retirements, organization policies and saved defaults. Restore a supported model without confusing a policy block with a code failure.

Genius Forges Editorial

Published · 3 min read

Jump to the guide ↓
A miniature wooden rail junction routing cobalt blocks away from a closed branch.
Original AI-generated conceptual illustration; not a product screenshot or observed test result.

At a glance

  • Use the final retirement notice for current replacement names.
  • A missing model can be an organization-policy issue.
  • Record the actual selected model before comparing task results.

What changed on October 19

Check team defaults and model policy before troubleshooting prompts. GitHub’s October 19 notice retires selected models across most Copilot experiences and lists supported alternatives. It also says Enterprise administrators may need to enable alternatives in model policies. Use the final notice, rather than relying only on the earlier announcement.

This is a migration of tool configuration and access. It does not tell you that a replacement will produce identical answers or that every editor surface updates at the same moment. The immediate objective is restoring a known, supported choice and documenting who can use it.

Map the affected selections

The final notice provides this replacement table:

Retired selection Suggested alternative
Gemini 3.7 Flash Gemini 3.8 Flash
GPT-5.5 or GPT-5.4 GPT-6.1 Sol
GPT-5.4 mini or GPT-5 mini GPT-6 Luna
Grok 4.5 Grok 4.7

Record where your team selected each model: editor settings, onboarding instructions, saved tasks or documented workflow examples. Search your own configuration files for exact retired names, but inspect each match before changing it. A historical article or audit record should keep its original model name.

Separate access failures from task failures

Ask an affected user to inspect the model selector in their current client. Have the administrator confirm that the intended replacement is enabled under organization policy. GitHub specifically points to checking availability in VS Code and on github.com after enabling access.

A missing model, a request rejected before inference and a model returning a wrong patch are three different failures. Log them separately. Otherwise a team can waste time rewriting prompts when the account is simply unable to select the configured model.

Update one workflow and keep a rollback record

Choose a low-impact internal task, such as explaining a small function or proposing a documentation correction. Record the old default, new selection, client version and date. Change the intended configuration only; do not broadly replace every occurrence of a provider name across a repository.

Use this brief to prepare a migration worksheet:

Review these configuration excerpts and onboarding notes:
{{redacted_configuration_and_notes}}
Retirement source: {{official_notice_text}}
Return location | retired selection | supported alternative |
policy owner | manual verification. Preserve historical records.
Do not infer that a model is enabled merely because it exists.
Flag unknown editor or organization settings for review.

The expected result is a small list of actual settings and owners, not a patch that invents configuration keys. Make changes through the configuration mechanism your client and organization support.

Verify with an observable result

Open a fresh Copilot conversation, confirm the selected model and run the bounded task. Check the answer against the source file. If code changes are proposed, inspect the diff and use the project’s relevant checks before adopting them. Save the model selection with the result so later reports are comparable.

GitHub says no action is required to remove the deprecated models. Do not add cleanup work that the platform already handles. Focus on alternatives, team access and accurate documentation.

Make the next retirement easier

Keep one short model inventory with a responsible owner and a link to the provider’s changelog. Include configuration defaults and account access separately. An approved fallback should name the work it is suitable for; a faster model may be fine for explanations but require a separate decision for a complex migration.

Use the existing model-update regression checklist when assessing behavior after access is restored. This article’s completion check is narrower: team members can select the intended supported model, and their saved defaults no longer point to a retired choice.

Sources and editorial review

Reviewed on 22 October 2026. Primary-source reporting and editorial analysis; the exercises and example results are illustrative, not hands-on performance measurements. Product access, regional rollout and commercial terms can change.

01 Choose a tool: compare the fit. 02 Use a prompt: start with context. 03 Check the result: keep your judgment.

About the author

Genius Forges Editorial

Editorial team

Practical AI reporting and guides grounded in primary sources, original analysis and clearly identified examples.

Did this help you complete your task?

Tell us what needs improving. No account or private output required.

Was this useful?
Report incorrect information or a broken link

Review evidence

Documentation-based review · 2026-10-22

Primary sources checked on 22 October 2026. News-led editorial selection, with access limitations recorded in the batch research notes. No measured exact-query popularity, comparative testing or traffic outcomes are claimed.