GitHub has set two Copilot dates three days apart. On October 19 six models leave every Copilot experience. On October 22 a new policy starts deciding, for Business and Enterprise customers, whether eligible features nobody has configured are on or off. Unless an administrator acts, the answer is on.

On September 13 this desk covered the previous wave, four models leaving on October 2.

Six models, October 19

GitHub's September 18 changelog post:

We will deprecate the following models across all GitHub Copilot experiences (including Copilot Chat, inline edits, ask and agent modes, and code completions) on October 19th, 2026

Leaving Copilot October 19	GitHub's suggested alternative
Gemini 3.7 Flash	Gemini 3.8 Flash
GPT-5.5	GPT-5.6 Sol
GPT-5.4	GPT-5.6 Sol
GPT-5.4 mini	GPT-5.6 Luna
GPT-5 mini	GPT-5.6 Luna
Grok 4.5	Grok 4.6

GPT-5.5 leaves OpenAI's own ChatGPT, ChatGPT Work and Codex five days earlier, on October 14. OpenAI's page says "This retirement does not apply to the OpenAI API." It points Codex users on Plus, Pro, Business, Enterprise and Edu plans to GPT-6 Sol; GitHub points to GPT-5.6 Sol.

The wording changed between waves

The October 2 notice said administrators "may need to enable access to the alternative models through their model policies in Copilot settings." The October 19 notice says they now switch on by themselves:

Under default model enablement, the suggested alternatives are automatically enabled for Copilot Enterprise and Copilot Business customers unless an administrator has turned off the global default or explicitly disabled the model.

GitHub's docs: "The models policy is already active." Some models start disabled whatever the default: pre-GA models, open-weight models (DeepSeek, Kimi K2.7 Code, Kimi K3), and Claude Fable 5 and 5.1. Kimi K3 is the October 2 wave's alternative for Kimi K2.7 Code, so an administrator has to enable it.

October 22: the feature default

The September 24 post adds a "Default policy for new features" setting: Enabled, Disabled, or Let organizations decide. "For the next 28 days, you can configure this policy, but it won’t affect feature access for your users yet." Then: "Starting October 22, the policy takes effect." The docs state the default plainly:

This policy is enabled by default. If you don't take action, unconfigured features will be enabled on October 22.

It covers the Features & clients page plus the Copilot code review and MCP servers policies, and it reaches forward: new GA features and features that move from preview to GA. "Explicit decisions are preserved." "Preview features remain opt-in." Exempt: the two restrictive model policies on GHE.com (data residency and FedRAMP), and storing local sessions in the cloud for Copilot CLI and VS Code.

The Fable exemption has an end date

A footnote on GitHub's supported-models page:

Customers can request to use Claude Fable 5 or Claude Fable 5.1 with zero data retention (ZDR) through the end of 2026 under a time-bound exemption while Anthropic rolls out Enterprise Frontier Safeguards (EFS). After that point, continued use of these models would require EFS, which will enable eligible customers to keep their data under their own control while also enabling automated safety monitoring.

Anthropic's own data retention page says these models "require 30-day data retention and are not available under ZDR unless expressly authorized by Anthropic." It gives no end-2026 date and does not mention EFS.

GitHub's own table has not caught up

The docs page's table of models "retired or scheduled for retirement from Copilot", as read September 27, lists none of the six October 19 retirements. It still names GPT-5.5 as the alternative for four retired models (GPT-5.2, GPT-4.1, GPT-5 and o3), and GPT-5 mini for four more (Grok Code Fast 1, o1-mini, o3-mini and o4-mini).

What administrators can do

Before October 19: update workflows and integrations that name any of the six. If you turned off the global model default, enable the alternatives yourself.
Before October 22: choose the default under AI Controls, then Copilot, or configure each feature explicitly. GitHub says a banner shows how many eligible policies are still unconfigured.
Before the end of 2026: enterprises using Fable under the exemption should ask what EFS will require; the footnote routes eligibility questions to the GitHub account team.

What the record does not say

Why the six are leaving.
Whether they stay available outside Copilot, beyond OpenAI's statement about GPT-5.5 on its API.
Which features a given enterprise has left Unconfigured; only its own settings show that.
What EFS costs, requires, or when it arrives.
When the end-2026 term was written. The footnote is undated.

GitHub Copilot drops six models on October 19, GPT-5.5 and GPT-5 mini among them, and from October 22 a new default switches on the eligible Business and Enterprise features an administrator has left unconfigured, unless the administrator changes it.