Short answer: GitHub will retire six Copilot models on October 19, 2026 across Copilot Chat, inline edits, ask and agent modes, and code completions. Replace Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini, and Grok 4.5 before the cutoff. GitHub has published a replacement for each model.[1]
Do this now
- Search workflows, custom agents, IDE settings, team documentation, and integrations for the six retiring model names.
- Enable the replacement models in Copilot policies if your organization does not use GitHub’s default model enablement.
- Change pinned model selections and test representative coding tasks.
- Tell users which replacement to select before October 19.
- Keep a rollback note for prompt or workflow changes, not for the retired model: GitHub will remove it from Copilot.
GitHub Copilot model replacement table
| Retiring model | GitHub’s suggested replacement | Deprecation date |
|---|---|---|
| Gemini 3.7 Flash | Gemini 3.8 Flash | October 19, 2026 |
| GPT-5.5 | GPT-5.6 Sol | October 19, 2026 |
| GPT-5.4 | GPT-5.6 Sol | October 19, 2026 |
| GPT-5.4 mini | GPT-5.6 Luna | October 19, 2026 |
| GPT-5 mini | GPT-5.6 Luna | October 19, 2026 |
| Grok 4.5 | Grok 4.6 | October 19, 2026 |
Who needs to take action?
Individual users: check the model selected in Copilot Chat and, separately, the model used for inline suggestions. GitHub documents these as distinct settings; changing the Chat model does not change the inline-suggestion model.[4][5]
Copilot Business and Enterprise administrators: review model policies before the deadline. GitHub says the suggested alternatives are automatically enabled when default model enablement is in use, unless an administrator has disabled the global default or explicitly disabled the replacement model.[1] Organization and enterprise owners can enable or disable model access for members with Business or Enterprise seats.[3]
Teams with custom workflows or integrations: search for hard-coded model names even if users normally rely on the visual model picker. A pinned identifier in an integration, runbook, prompt-evaluation suite, or custom agent may be easy to miss during a UI-only review.
Migration checklist for administrators
1. Inventory every model reference
Search for the exact retiring names in repositories, workflow configuration, custom-agent files, internal setup guides, screenshots, onboarding documents, and test fixtures:
Gemini 3.7 Flash
GPT-5.5
GPT-5.4
GPT-5.4 mini
GPT-5 mini
Grok 4.5Do not assume every reference is machine-readable. A team wiki telling developers to choose GPT-5.5 manually can create just as much confusion as a pinned configuration value.
2. Confirm replacement-model access
Check your Copilot organization or enterprise settings and verify that the appropriate successor models are allowed. This step matters most when the global default has been turned off or model policies are managed individually. GitHub’s documentation explains that model availability depends on the user’s plan and on organization or enterprise policy.[2][3]
3. Update pinned selections
Use GitHub’s published mapping as the starting point, but treat the change like a dependency upgrade rather than a text-only rename. Update one workflow or team at a time where practical, then run the same evaluation tasks against the old and new selections before the cutoff.
4. Test behavior, not just availability
- Run a representative code-completion task.
- Run a repository-aware question in Copilot Chat.
- Test an edit or refactor that uses project context.
- Exercise an agent-mode task that calls tools.
- Compare correctness, unwanted file changes, response time, and review effort.
- Repeat any high-risk test involving deployment, security, permissions, or data migration.
A replacement appearing in the picker proves access, not workflow parity. Keep the test set small, repeatable, and based on real repository work.
5. Update team guidance
Publish a one-line mapping in the place developers already use for engineering updates. Include the October 19 deadline and identify who owns Copilot policy changes. Remove stale screenshots and model recommendations after migration.
How to change the Copilot Chat model
On GitHub.com, open Copilot Chat and use the current-model dropdown at the bottom of the chat to select another available model. GitHub notes that Business users need their organization to permit model switching.[4]
In supported IDEs, use the model selector shown in Copilot Chat. If a replacement is missing, first check your plan, client support, extension version, and organization policy against GitHub’s current supported-model table.[2]
How to change the inline-suggestion model in VS Code
- Update VS Code and the GitHub Copilot extension.
- Open the Command Palette with
Ctrl+Shift+Pon Windows/Linux orCommand+Shift+Pon macOS. - Run GitHub Copilot: Change Completions Model.
- Select an available replacement.
- Open Settings and search for
copilot completionto confirm the selected completion model.
These steps come from GitHub’s current inline-suggestion model guide. Availability can vary, and the picker may show only one model when no alternative is offered.[5]
What happens on October 19?
GitHub says the six models will be deprecated across all Copilot experiences covered by the announcement. Users do not need to remove the retired models after deprecation; GitHub handles their removal. The action required before the date is to update workflows and integrations and ensure replacements are enabled where policy requires it.[1]
FAQ
Which GitHub Copilot models are retiring on October 19, 2026?
Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini, and Grok 4.5.[1]
Will GitHub automatically enable the replacements?
For Copilot Business and Enterprise customers using default model enablement, GitHub says the suggested alternatives are automatically enabled unless an administrator disabled the global default or explicitly disabled that model. Administrators with restrictive policies should verify access manually.[1]
Does changing the Chat model also change code completions?
No. GitHub documents Chat and inline-suggestion model selection separately, and explicitly says changing the Chat model does not affect inline suggestions.[4]
Do I need to delete retired models after the deadline?
No. GitHub says no action is required to remove them after deprecation.[1]