Skip to content
HUIAHUB

News

GitHub Copilot deprecates four models: what replaces each one and where

GitHub removed Gemini 3.5 and 3.6 Flash, Kimi K2.7 Code and Claude Opus 4.7 from Copilot on October 2, 2026. Replacements, plan limits and admin steps.

Edited by Published 4 min read3 sources

GitHub deprecated four models across GitHub Copilot on October 2, 2026: Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code and Claude Opus 4.7. The change took effect the day it was announced and covers Copilot Chat, inline edits, ask and agent modes, and code completions. GitHub names a replacement for each model and says users do not need to do anything to remove the old ones.

Which Copilot models are gone and what replaces them

The GitHub changelog entry (opens in a new tab) gives one date, 2026-10-02, for all four models, with these suggested alternatives:

Deprecated modelSuggested alternative
Gemini 3.5 FlashGemini 3.8 Flash
Gemini 3.6 FlashGemini 3.8 Flash
Kimi K2.7 CodeKimi K3
Claude Opus 4.7Claude Opus 5.5

GitHub's supported models reference (opens in a new tab) now lists all four in its model retirement history with a retirement date of 2026-10-02. There, the suggested alternatives are worded more generally: "Latest Claude Opus model", "Latest Gemini Flash model" and "Latest Kimi model". The changelog uses the word deprecated, but the docs file them as retired, and neither page mentions a grace period.

What Copilot users need to do

Individual users mostly need to check saved choices. If a workflow, prompt file, custom agent or integration names one of the four models, it should point to a supported model instead. GitHub's note says the deprecated models disappear without any action on your part.

Copilot Enterprise administrators have a specific task. The replacement models may be switched off by model policy, so admins may need to enable them in Copilot settings. GitHub suggests checking your own Copilot settings to confirm the policy is on for each model. Once a policy is enabled, the model shows up in the Copilot Chat model picker in VS Code and on github.com. Enterprise customers with questions are pointed to their account manager.

Where the replacement models are available

The supported models reference shows some differences between the old and new options that matter in practice.

Claude Opus 5.5 is marked GA. The plan table lists it for Copilot Pro+, Max, Business and Enterprise, but not for Copilot Pro. A Copilot Pro user who relied on Opus 4.7 therefore has no direct Opus replacement on that plan, based on the current table. Opus 5.5 is listed in every client in the per-client table, from GitHub.com to Xcode and JetBrains IDEs.

Gemini 3.8 Flash is also GA and is listed on all five plans. The per-client table marks it as available in Copilot CLI, VS Code, Visual Studio, Eclipse, Xcode and JetBrains IDEs, but not on GitHub.com. That sits awkwardly with the changelog's general line about models appearing on github.com once enabled.

Kimi K3 is GA, on all five plans and in every listed client. Kimi K3 and Claude Opus 5.5 both support the 1 million token context window and configurable reasoning levels. Gemini 3.8 Flash is not in that extended-capabilities table. GitHub notes that the 1 million token window works only in VS Code and Copilot CLI.

For IDE versions, the docs list Visual Studio 17.14.22 or 18.1.0 for Gemini 3.8 Flash, Visual Studio 17.14.6 for Claude Opus 5.5, and VS Code v1.131 for Kimi K3. Other entries are marked TBD.

What the replacements cost

Copilot bills usage above plan allowances in AI credits, where 1 AI credit is $0.01 USD. GitHub's models and pricing page (opens in a new tab) lists these rates per 1 million tokens:

  • Claude Opus 5.5: $4.00 input, $0.20 cached input, $5.00 cache write, $20.00 output.
  • Kimi K3: $3.00 input, $0.30 cached input, $15.00 output.
  • Gemini 3.8 Flash: $0.75 input, $0.075 cached input, $3.75 output, a promotional price that runs through December 31, 2026.

The pricing page does not list the four retired models. That leaves no official side-by-side to show whether moving from Opus 4.7 to Opus 5.5, or from Kimi K2.7 Code to Kimi K3, costs more per token.

Open questions

  • GitHub gives no reason for the deprecations.
  • No advance notice date appears in the changelog. The retirement date and the announcement date are the same.
  • The changelog does not say what happens to a Copilot request, or to an automation, that still names a removed model.
  • Gemini 3.8 Flash's availability on GitHub.com is unclear, given the gap between the changelog wording and the per-client table.

Frequently asked questions

Will Copilot keep using a retired model if it was set as my default?

GitHub says no action is needed to remove the deprecated models, which suggests they drop out of model pickers on their own. It does not say which model a saved preference falls back to. Check your model selection in each client after the change.

Does Auto model selection pick up the replacements?

The Auto model selection table in GitHub's docs lists Claude Opus 5.5 but does not list Gemini 3.8 Flash or Kimi K3. If you rely on Auto, the Gemini and Kimi replacements need to be chosen by hand. Available Auto models can also be narrowed by your organization's model policies.

Where can I give feedback on the change?

GitHub points users to the Copilot conversations category in the GitHub Community discussions. Enterprise customers can also go through their account manager, according to the changelog.

Sources

  1. 1.Selected models in GitHub Copilot deprecated - GitHub ChangelogOfficial docsPrimarygithub.blog
  2. 2.docs.github.com/en/copilot/reference/ai-models/supported-modelsOfficial docsPrimary
  3. 3.docs.github.com/en/copilot/reference/copilot-billing/models-and-pricingOfficial docsPrimary

Spotted an error? See our corrections policy.

Enjoyed this? Get the weekly brief.

New guides, Project & Tool of the Week and the AI developments that matter. One email, no spam.

Read next