GitHub gave Copilot administrators a new automation interface on the same day it retired six models across most Copilot experiences. The October 19 announcements put model migration and policy management on the same operational calendar. But the distinction matters: configuring a model policy does not establish that a particular developer can use that model. GitHub’s own documentation warns against treating the returned policy state as a final access answer.
The retirement list needs a date
The October 19 retirement notice names six outgoing models. It suggests Gemini 3.8 Flash for Gemini 3.7 Flash; GPT-6.1 Sol for GPT-5.5 and GPT-5.4; GPT-6 Luna for GPT-5.4 mini and GPT-5 mini; and Grok 4.7 for Grok 4.5.
Those are suggested replacements, not demonstrated equivalents. The notice covers most experiences, including chat, inline edits, ask and agent modes, and code completions. It says administrators need no action to remove the retired models, but Enterprise administrators may need to enable alternatives through model policies.
The September 18 advance notice listed the same retirement date and outgoing models, but some different alternatives: GPT-5.6 Sol, GPT-5.6 Luna, and Grok 4.6. The notices do not explain the change. A migration record should therefore preserve which dated recommendation informed the decision.
What the API actually controls
GitHub’s general availability announcement introduces GET and PATCH REST endpoints for model policies. It names enterprise, enterprise-team, enterprise-app, and organization scopes. That gives administrators a programmatic way to read and change policy settings. The announcement does not describe those operations as a per-user access test.
The enterprise AI controls reference documents enterprise and enterprise-team operations, including policy values such as enabled, disabled, and optional. It also specifies the corresponding read and write permissions. These values describe configured policy; they should not be translated directly into a claim that every affected developer has access.
Delegation changes the access question
GitHub’s API documentation explicitly cautions that returned policy states do not necessarily determine whether an individual user can access a model. A successful policy read therefore answers a narrower question than an administrator may initially ask.
The enterprise model-availability documentation describes a baseline that can enable, disable, or delegate model decisions to organizations or enterprise teams. It identifies Enterprise owners and Copilot Business or Enterprise as eligible for these controls.
Enterprise-teams mode is an opt-in preview, and organization-level model settings are deactivated in that mode. A model disabled at enterprise level cannot be enabled for a team. Conversely, a model enabled by any of a user’s enterprise teams is available to that user under the documented rules. Reading a baseline without understanding the delegation mode can therefore leave the access picture incomplete.
A migration check with explicit limits
A practical review should inventory named model references, record the dated replacement guidance, and read the applicable policy state. Administrators should then establish the intended baseline and delegated rules before checking availability in the Copilot surface their developers actually use.
Task checks belong in that proposed evaluation too: a suggested replacement should be assessed against the work it must perform. No independent tenant or task testing was conducted for this report.
As with GitHub’s dependency-license enforcement controls, the useful question is what a control establishes. Here, the API makes policy administration repeatable. It does not, by itself, settle individual access or replacement-model behavior.