GHE users can't see/use all enabled models

[jochenjonc]

🏷️ Discussion Type

Question

💬 Feature/Topic Area

Copilot Enterprise

Body

My company recently switched from a GitHub organisation to GHE and we use this mainly to manage Copilot licenses. But since the switch we can no longer use the latest models like OpenAI GPT-5.6 Luna even if they are enabled just like any other model.

Can anyone help me enabling these models? Or explain the reason?

[dinildamsith]

It might be a GitHub Enterprise setting playing hide-and-seek 😄

First, check Settings → Copilot → Policies and make sure the models you want are enabled for your organization. Model availability can also depend on your Copilot plan and enterprise-level policies.

If the models are enabled but still don't show up, check whether the Copilot licenses are assigned correctly and whether an enterprise policy is overriding the organization settings.

Also, are you using GitHub Enterprise Cloud or Enterprise Server? That could help narrow down what’s happening.

[antoniorotundo2]

Since the move from an org to GHE is the detail that matters most here, the likely cause is not a per-model setting but a change in which layer of policy actually governs access. On a standalone organization, the org's own model policy is authoritative. Once the org sits under an enterprise account, GitHub Docs describe enterprise-level model policy as able to override what the organization shows: a model can appear "enabled" at the org level while the enterprise policy has it as unconfigured or restricted, and the org-level setting only takes effect if the enterprise has left that specific model as "optional" for organizations to decide.

There is also a change that shipped very recently and lines up with the timing you describe. GitHub rolled out a preview called enterprise teams mode for Copilot model policies, starting August 3, 2026. When an enterprise opts into it, organization-level model settings stop applying entirely, and access to each model is controlled only through enterprise-level policy and enterprise teams, with a least-restrictive rule where a user gets a model if any one of their enterprise teams grants it. If your GHE admin opted into that preview around when the switch happened, that alone would explain org-level "enabled" no longer being sufficient, since the org toggle stops being read.

Worth checking, in order: on the enterprise account, under Copilot, whether Enterprise teams mode is turned on, since that changes where the real switch lives; if it is on, the model has to be granted through an enterprise team rather than the organization; if it is off, check the enterprise's own model list under Copilot > Configure models, since a model left as Disabled or simply unconfigured at the enterprise level blocks it regardless of the organization's setting, and newer or preview models specifically are not auto-enabled by default even when a general "default availability" policy is on. If neither explains it, whether your Copilot plan is Business or Enterprise also matters, since some models are plan-gated independently of any policy toggle.

If my answer was helpful to you, please consider marking it as the correct answer, thanks.

[jochenjonc]

The Editor preview features were already enabled. I don't know what happend, but I didn't change anything and now I can't access any model. It's really frustrating.

[antoniorotundo2]

That is a different failure than before, going from one model family not showing up to nothing working with no change on your end points away from a policy detail and toward either an account-level state or a platform-side issue. I checked githubstatus.com just now and there is no currently listed incident matching this specifically, so I cannot confirm it is a known outage, but it is worth checking that page yourself in real time since status can change quickly and a component-specific degradation does not always match what a general search turns up.

Two things worth trying that are quick and reversible. Enterprise teams mode is a very new preview, it started rolling out on August 3, so it has had about a week in the wild, and GitHub's own docs note that during the preview you can roll back to your previous model policy configuration if something breaks. If your enterprise admin has access to that rollback option, trying it and seeing whether access returns would tell you whether the preview itself is the source rather than something specific to the models you listed. Separately, a full sign-out and sign-in in the IDE, and confirming the Copilot extension is on its latest version, rules out a stale session token, which is a common and boring cause of "everything just stopped" with no configuration change.

If neither restores access, this has moved past what troubleshooting in this thread can resolve, since going from one degraded model to a total loss of access on an enterprise account, with nothing changed on your side, is exactly the kind of thing that needs someone with visibility into your specific tenant. A Support ticket at this point, mentioning that Editor preview features was already on, that Enterprise teams mode is in use, and the exact time it stopped working, gives them what they would need to look at your account directly rather than guess from outside.

[antoniorotundo2]

Separate from the troubleshooting above, if you need GPT-5.6 models sooner than IT can resolve the enterprise-side issue, Zed supports it directly, either through Zed's own hosted plans or by adding your own OpenAI API key under language_models in settings.json. Worth flagging that this would mean routing prompts and code outside your company's managed Copilot licensing, so it is a personal stopgap to weigh against your org's policies, not a fix for the underlying issue.

[jochenjonc]

I have already created 2 support tickets because it started happening a week ago.

[antoniorotundo2]

Worth flagging since it did not come up before: "started happening a week ago" lines up closely with August 3, when Enterprise teams mode for Copilot model policies began rolling out as a preview. If your enterprise opted into it around then, that timing is not a coincidence, it would explain both the original single-model gap and the later drop to zero as the same rollout, not two separate issues.

If you have the ticket numbers for the two Support cases already open, adding that specific date correlation to one of them, along with when your admin actually opted into Enterprise teams mode if you can confirm it, gives Support a concrete lead to check server-side rather than starting from "it stopped working" with no anchor point. Worth keeping this to one ticket rather than two running in parallel too, since a single thread with all the details tends to move faster than duplicate reports split across two.

[ibrexpertdev]

🏷️ Discussion Type

Question

💬 Feature/Topic Area

Copilot Enterprise

Body

My company recently switched from a GitHub organisation to GHE and we use this mainly to manage Copilot licenses. But since the switch we can no longer use the latest models like OpenAI GPT-5.6 Luna even if they are enabled just like any other model.

Can anyone help me enabling these models? Or explain the reason?

github playing hide-and-seek 😂