GitHub Copilot Deprecates Claude Opus 4.7, Gemini 3.5 and 3.6 Flash and Kimi K2.7 Code
On October 2, 2026, GitHub deprecated four AI models across GitHub Copilot: Claude Opus 4.7, Gemini 3.5 Flash, Gemini 3.6 Flash and Kimi K2.7 Code. GitHub names Claude Opus 5.5, Gemini 3.8 Flash and Kimi K3 as the replacements. Developers do not have to remove anything, but administrators on Copilot Business and Copilot Enterprise may need to switch the replacement models on before their teams can select them.
For a business, this is a small change with a practical side: the tool your developers use every day may now be answering with a different model than it did last week. If your team builds or maintains software with AI coding tools, or you rely on an outside team for software development services, it is worth ten minutes to confirm that nothing was left pointing at a model that is gone.
What changed in GitHub Copilot on October 2, 2026?
GitHub deprecated four models on October 2, 2026, and the change took effect the same day. The GitHub changelog entry says the models were deprecated "across all GitHub Copilot experiences (including Copilot Chat, inline edits, ask and agent modes, and code completions)".
The four models and the alternatives GitHub suggests are:
- Claude Opus 4.7, replaced by Claude Opus 5.5
- Gemini 3.5 Flash, replaced by Gemini 3.8 Flash
- Gemini 3.6 Flash, replaced by Gemini 3.8 Flash
- Kimi K2.7 Code, replaced by Kimi K3
As of October 6, 2026, none of the four deprecated models appears in GitHub's list of supported Copilot models, while all three replacements do.
Who is affected by the Copilot model deprecations?
Anyone who had picked one of the four models in Copilot is affected, along with the administrators who control which models their organization may use. That covers individual developers on paid plans as well as companies on Copilot Business and Copilot Enterprise.
The effect is different for each group. A developer who chose Claude Opus 4.7 in the model selector will need to choose another model. An administrator has a separate job, described below, because in organizations a model has to be allowed by policy before it shows up for anyone.
Business owners who do not use Copilot themselves are affected indirectly. If your in-house developers or your agency use Copilot on your codebase, the model behind their suggestions and code reviews may have changed on October 2, 2026.
Is Claude Opus 4.7 being retired everywhere?
No. The deprecation applies to GitHub Copilot, not to the model itself. Anthropic's own model deprecations page lists claude-opus-4-7 as Active, with a retirement date "not sooner than April 16, 2027", as of October 6, 2026.
This distinction matters if your company uses AI in two places. A developer tool such as Copilot keeps its own, shorter list of models and changes it on its own schedule. An application you have built that calls a model directly through a provider's API follows that provider's schedule instead. Anthropic notes on the same page that partner-operated platforms set their own retirement schedules, so dates can differ from one platform to another.
In plain terms: if your product calls Claude Opus 4.7 through Anthropic's API, the October 2 Copilot change does not switch it off. If your developers used Claude Opus 4.7 inside Copilot, that option is the one that changed.
What do Copilot administrators need to do?
Administrators need to make sure the replacement models are enabled in their Copilot model policies; nothing has to be done to remove the old ones. GitHub's changelog says Copilot Enterprise administrators may need to enable access to the alternative models through model policies in Copilot settings, and that "no action is required to remove the deprecated models".
GitHub also describes how to confirm the result. An administrator can check individual Copilot settings to see that the policy for a specific model is enabled. Once it is, the model appears in the model selector in VS Code and on github.com.
If a replacement is not enabled, developers will not see that model in their list, which tends to produce a support request along the lines of "Opus disappeared". Checking the policy first saves that round trip.
Will the replacement models behave the same way?
No, a replacement model is a different model, so its output will differ even on the same prompt. GitHub's changelog names the alternatives but makes no statement about how their behavior compares, and neither should anyone who has not tested them on your code.
What tends to change
Three things usually shift when a coding model is swapped: the style and length of the code it writes, how it follows written team conventions and custom instructions, and how much of your Copilot usage allowance each request consumes. The first two affect code review. The third affects your bill or your monthly limits, so check the usage and billing information for the replacement model under your own plan rather than assuming it matches the old one.
Automations are the part people forget
Interactive use is easy to fix, because a developer sees the missing model and picks another. Scripts, saved agent configurations and internal documentation that name a specific model are different. Nobody is watching them, so a reference to a deprecated model can sit unnoticed until a job fails or quietly falls back to something else.
What else did GitHub change for Copilot on the same day?
GitHub also announced on October 2, 2026 that Copilot code review can now be requested through its REST and GraphQL APIs, and confirmed that Balanced is the default review effort level. According to the code review changelog entry, this applies to the Copilot Pro, Pro+, Max, Business and Enterprise plans, and a request made through the API can set the review effort level.
For a business this means automated code review can now be triggered from your own tooling, such as a deployment pipeline, instead of only from the GitHub interface. It is a separate change from the model deprecations, but teams reviewing their Copilot setup this week can look at both together.
What should a business do now?
A business should spend a short, fixed amount of time confirming that its Copilot setup still matches what the team expects. The following checklist is enough for most companies:
- Ask whoever administers Copilot to confirm that Claude Opus 5.5, Gemini 3.8 Flash and Kimi K3 are enabled in your model policies, if your team wants them.
- Ask developers which model they had selected before October 2, 2026 and which one they are using now.
- Search scripts, agent configurations, onboarding documents and internal guides for the four deprecated model names and update them.
- Check how the replacement model counts against your plan's usage allowance.
- Review the next few AI-assisted pull requests a little more carefully than usual, since the model writing them has changed.
- If you use an outside development team, ask them the same questions about work done on your code.
Model lists in developer tools keep changing as providers release new versions. The lasting fix is to treat the model name as a setting that someone owns, the same way a company tracks a framework version, rather than something each developer chose once and forgot. If your company is also building software that uses AI models directly, our explainer on what an AI agent is covers how those systems depend on a specific model.
Quick answers
Which models did GitHub Copilot deprecate on October 2, 2026?
GitHub deprecated Claude Opus 4.7, Gemini 3.5 Flash, Gemini 3.6 Flash and Kimi K2.7 Code across all GitHub Copilot experiences on October 2, 2026.
What replaces Claude Opus 4.7 in GitHub Copilot?
GitHub names Claude Opus 5.5 as the suggested alternative to Claude Opus 4.7 in Copilot. Gemini 3.8 Flash replaces Gemini 3.5 Flash and Gemini 3.6 Flash, and Kimi K3 replaces Kimi K2.7 Code.
Do I need to do anything about the GitHub Copilot model deprecations?
You do not need to remove the deprecated models. Copilot Enterprise administrators may need to enable the replacement models in their Copilot model policies, and teams should update any scripts or documents that name the old models.
Is Claude Opus 4.7 still available outside GitHub Copilot?
Yes. As of October 6, 2026, Anthropic lists Claude Opus 4.7 as Active on its own API, with retirement not sooner than April 16, 2027. The October 2, 2026 change applies only to GitHub Copilot.
Why can my developers not see the new model in Copilot?
In an organization, a Copilot model only appears in the model selector after an administrator enables it in the model policy. If Claude Opus 5.5, Gemini 3.8 Flash or Kimi K3 is missing, ask your Copilot administrator to check that policy.
Conclusion
The October 2, 2026 deprecations remove four older models from GitHub Copilot and point teams to Claude Opus 5.5, Gemini 3.8 Flash and Kimi K3. Nothing breaks for most developers, but administrators should confirm the replacements are enabled, and someone should check scripts and documents for the old names. The change is specific to Copilot; applications that call these models through a provider's own API follow that provider's dates.
Entrant Technologies builds websites, web applications, mobile apps and custom software. If you are unsure how a model change affects the software you run, you can contact us and we will take a look.