Quick answer: GitHub will retire Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code, and Claude Opus 4.7 from all GitHub Copilot experiences on October 2, 2026. Replace them with Gemini 3.8 Flash, Kimi K3, and Claude Opus 5 respectively. Business and Enterprise administrators should also confirm that the replacement models are enabled in Copilot model policies before users switch.[1]
The migration table
| Retiring Copilot model | Retirement date | GitHub’s suggested replacement | Action |
|---|---|---|---|
| Gemini 3.5 Flash | October 2, 2026 | Gemini 3.8 Flash | Enable the replacement, test your normal tasks, then change saved selections. |
| Gemini 3.6 Flash | October 2, 2026 | Gemini 3.8 Flash | Move pinned chat, agent, and completion workflows to Gemini 3.8 Flash. |
| Kimi K2.7 Code | October 2, 2026 | Kimi K3 | Test for behavior and cost differences before making Kimi K3 the default. |
| Claude Opus 4.7 | October 2, 2026 | Claude Opus 5 | Enable Opus 5, run a representative evaluation, and update pinned selections. |
GitHub says the retirement covers Copilot Chat, inline edits, Ask and Agent modes, and code completions. You do not have to manually remove the old models after retirement, but you should update workflows and integrations that explicitly select them.[1]
Do this now: seven-step checklist
- Identify your retiring model. Check the current model in Copilot Chat and the separate completion-model setting in each IDE you use.[4][6]
- Ask an administrator to enable the replacement. This is especially important for Copilot Business and Enterprise accounts.[1][5]
- Update your IDE and Copilot extension. Recent models can require newer clients; GitHub’s current support table lists VS Code 1.131 for Kimi K3 and 1.128 for Claude Opus 5, while the minimum VS Code version for Gemini 3.8 Flash is still listed as TBD.[3]
- Switch Chat and inline completions separately. Changing the Copilot Chat model does not change the model used for inline suggestions.[4][6]
- Run the same small evaluation on the old and new model. Use the test pack below before the old model disappears.
- Search repositories and team documentation for pinned model names. Replace only confirmed Copilot references; do not blindly rename unrelated Gemini, Kimi, or Claude API configurations.
- Confirm the model picker on every surface your team uses. Availability depends on the Copilot plan and client, and GitHub says rollout for Gemini 3.8 Flash is gradual.[2][3]
How to switch the model in Copilot Chat
GitHub.com
- Open Copilot Chat on GitHub.[4]
- Select the CURRENT-MODEL menu at the bottom of Chat.[4]
- Choose Gemini 3.8 Flash, Kimi K3, Claude Opus 5, or another approved supported model.[3][4]
- Send a test prompt and confirm the selected model remains visible.
GitHub also lets you regenerate a response with another model while keeping the conversation context.[4]
Visual Studio Code
- Open Copilot Chat from the VS Code title bar.[4]
- Select CURRENT-MODEL at the bottom of the Chat view.[4]
- Choose the replacement model.[4]
- Start a new chat if you were using Auto model selection and need to switch to a named model.[4]
If the replacement is missing, update VS Code and the Copilot extension, confirm plan eligibility, and ask your administrator to inspect the model policy. Gemini 3.8 Flash is available to Copilot Pro, Pro+, Max, Business, and Enterprise users, but GitHub says its rollout is gradual.[2]
Visual Studio, JetBrains, Eclipse, and Xcode
In supported clients, open Copilot Chat and use the CURRENT-MODEL menu in the Chat panel.[4] GitHub documents model selection for Visual Studio, JetBrains IDEs, Eclipse, and Xcode, although the exact menu position differs by client.[4]
Change the inline-suggestion model separately
A common migration mistake is changing Chat and assuming code completions changed too. GitHub explicitly says the Copilot Chat selection does not affect inline suggestions.[4]
VS Code
- Press
Ctrl+Shift+Pon Windows/Linux orCommand+Shift+Pon macOS.[6] - Run GitHub Copilot: Change Completions Model.[6]
- Select an available replacement.[6]
- Open Settings, search for
copilot completion, and check GitHub > Copilot: Selected Completion Model. An empty field means the default model is being used.[6]
Visual Studio
Open Settings → Options → Copilot Completions, then select the model from the completions dropdown.[6] GitHub requires Visual Studio 17.14 Preview 2 or later for changing this model.[6]
JetBrains IDEs
Open Settings → Tools → GitHub Copilot → Completions, choose Model for completions, and save the setting.[6]
Not every Chat model is necessarily available for inline suggestions. The completion picker shows only alternatives currently supported for that client and account.[6]
Admin checklist for Copilot Business and Enterprise
Organization owners can manage model availability from the organization Copilot settings.[5]
- GitHub profile menu → Organizations.[5]
- Select the organization, then open Settings.[5]
- Under Code, planning, and automation, select Copilot.[5]
- Open Models.[5]
- Enable the replacement models required by the team.[5]
Enterprise policy can override an organization’s choice. If an enterprise owner has enforced a model policy, an organization owner cannot override it locally.[5]
Before declaring the migration complete, an administrator should:
- confirm Gemini 3.8 Flash, Kimi K3, or Claude Opus 5 is enabled as required;
- ask one user from each affected team to verify the model appears in the relevant picker;
- verify each team’s billed Copilot account or organization is the expected one;
- record the approved fallback model;
- review usage after switching, because the replacement can have a different token price.[3][7]
Find pinned model names in repositories
Run this from the root of a repository to find literal references in tracked configuration, scripts, workflow files, and documentation:
git grep -nE 'Gemini 3\.(5|6) Flash|gemini-3\.(5|6)-flash|Kimi K2\.7 Code|kimi-k2\.7|Claude Opus 4\.7|claude-opus-4\.7'
If git grep returns nothing, that means no matching text was found in tracked files—not that every user’s local IDE selection is safe. Check the UI settings separately.
Review every result before editing. A repository can contain direct provider API model IDs that are unrelated to GitHub Copilot. GitHub’s announcement is specifically about model availability inside Copilot; it does not announce retirement of the providers’ independent APIs.[1]
Use a simple migration register for a team:
| Owner | Surface | Old model | Replacement | Policy enabled | Test passed | Updated |
|---|---|---|---|---|---|---|
| [name] | VS Code Chat | Gemini 3.6 Flash | Gemini 3.8 Flash | Yes/No | Yes/No | [date] |
| [name] | Inline suggestions | [model] | [model] | Yes/No | Yes/No | [date] |
| [name] | Copilot cloud agent | Claude Opus 4.7 | Claude Opus 5 | Yes/No | Yes/No | [date] |
A five-prompt migration test pack
Run the same prompts against the current and replacement models using a small, non-sensitive test repository. Judge correctness and behavior rather than writing style alone.
1. Repository understanding
Explain the request flow from the public entry point to data persistence. Cite exact file paths and function names. Mark anything you cannot verify from the repository.
Check whether the model identifies real files, follows the correct call path, and avoids inventing symbols.
2. Focused bug fix
Find the cause of [known small bug]. Propose the smallest safe patch, add a regression test, and do not modify unrelated files.
Compare patch size, test quality, and whether the model respects scope.
3. Test execution
Run the narrowest relevant tests first. If they pass, run the affected package test suite. Report the exact commands, exit codes, and any failures without hiding them.
Check whether the model actually executes verification instead of merely describing it.
4. Security review
Review this change for authentication, authorization, injection, secret exposure, and unsafe file or network access. Report only findings supported by a file and line reference.
Reject vague warnings that do not map to the code.
5. Instruction following
Make no code changes. Produce a plan limited to five steps, identify the riskiest assumption, and list the one question that must be resolved before implementation.
This reveals whether the replacement follows constraints before you allow it to edit a live branch.
Score each test from 1 to 5 for correctness, groundedness, scope control, useful tool use, and latency. Use the same repository state and prompts for both models.
Replacement-specific notes
Gemini 3.5 or 3.6 Flash → Gemini 3.8 Flash
GitHub has made Gemini 3.8 Flash generally available across a broad set of Copilot clients, including VS Code, Visual Studio, Copilot CLI, the cloud agent, the Copilot app, JetBrains IDEs, Xcode, and Eclipse. GitHub describes the rollout as gradual and says Business and Enterprise administrators can control access through the model policy.[2]
GitHub’s current pricing page lists Gemini 3.8 Flash at promotional usage-based pricing through December 31, 2026. Do not assume that temporary pricing will continue after that date; check the live pricing page when planning budgets.[2][7]
Kimi K2.7 Code → Kimi K3
GitHub lists Kimi K3 as a supported GA Copilot model and names it as the suggested replacement for Kimi K2.7 Code.[1][3] The pricing table currently places Kimi K3 above Kimi K2.7 Code on input and output token prices, so teams using usage-based billing should test quality and monitor actual AI-credit use rather than treating the change as a name-only upgrade.[7]
Claude Opus 4.7 → Claude Opus 5
GitHub lists Claude Opus 5 as the suggested replacement and as a supported GA model. Its support table currently lists VS Code 1.128 and Visual Studio 17.14.22 as known minimum versions, with several other client minimums still marked TBD.[1][3]
Muddaser.com also has a separate Claude Opus 5 in GitHub Copilot enable and test guide for teams that need a model-specific walkthrough.
What happens if you do nothing?
GitHub says no action is needed to remove a model after it is deprecated.[1] However, a pinned workflow or integration can still need attention if it expects an old selection. The safer approach is to migrate while both models are available, test the replacement on representative work, and keep an approved fallback.
Do not promise users that every existing conversation will continue identically. Models differ, and model availability also depends on the plan and client.[3][4]
Troubleshooting
The replacement is not in the model picker
- Confirm the model is supported by your Copilot plan and client.[3]
- Update the IDE and Copilot extension.
- For Gemini 3.8 Flash, remember that rollout is gradual.[2]
- Ask a Business or Enterprise administrator to check Copilot → Models.[1][5]
- If enterprise policy controls the setting, escalate to the enterprise owner.[5]
Chat changed, but inline suggestions did not
That is expected when only the Chat picker was changed. Use the separate completion-model controls described above.[4][6]
The IDE is updated, but a model remains unavailable
Check the minimum-version table and the account’s Copilot plan. GitHub marks some client minimum versions as TBD, so the documentation may not yet provide a version-specific answer for every IDE.[3]
Usage cost changed after migration
The replacements do not all have the same token prices as the retiring models. Review GitHub’s live pricing table and compare actual AI-credit usage after the switch.[7]
Frequently asked questions
Which GitHub Copilot models retire on October 2, 2026?
Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code, and Claude Opus 4.7.[1]
What replaces Gemini 3.5 Flash and Gemini 3.6 Flash?
GitHub recommends Gemini 3.8 Flash for both models.[1]
What replaces Kimi K2.7 Code?
GitHub recommends Kimi K3.[1]
What replaces Claude Opus 4.7?
GitHub recommends Claude Opus 5.[1]
Does changing Copilot Chat also change inline completions?
No. GitHub documents Chat and inline suggestions as separate model selections.[4][6]
Do administrators need to remove the retired models?
No. GitHub says no action is required to remove them after deprecation. Administrators should instead enable and validate access to the replacements.[1]
Is Gemini 3.8 Flash available to Copilot Pro users?
Yes. GitHub lists Copilot Pro, Pro+, Max, Business, and Enterprise, but says rollout is gradual.[2]
Will the replacement cost the same?
Not necessarily. GitHub bills usage according to tokens and model-specific pricing, and the live table shows differences between several retiring models and their replacements.[7]
Sources
[1] https://github.blog/changelog/2026-09-03-upcoming-deprecation-of-selected-github-copilot-models — GitHub Copilot model deprecations — October 2, 2026
[2] https://github.blog/changelog/2026-09-03-gemini-3-8-flash-is-now-available-in-github-copilot — Gemini 3.8 Flash available in GitHub Copilot
[3] https://docs.github.com/en/copilot/reference/ai-models/supported-models — Supported AI models in GitHub Copilot
[4] https://docs.github.com/en/copilot/how-tos/use-ai-models/change-the-chat-model — Changing the AI model for GitHub Copilot Chat
[5] https://docs.github.com/copilot/managing-github-copilot-in-your-organization/managing-policies-and-features-for-copilot-in-your-organization — Managing Copilot models in an organization
[6] https://docs.github.com/en/copilot/how-tos/use-ai-models/change-the-completion-model — Changing the model for Copilot inline suggestions
[7] https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing — GitHub Copilot models and pricing