Note
GitHub Models for organizations and repositories is in public preview and subject to change.
About GitHub Models
GitHub Models automates, enhances, and streamlines AI-powered software development processes within GitHub. You can use GitHub Models to manage and optimize prompts, compare models, and create robust evaluations. See About GitHub Models.
Prerequisites
For GitHub Models to be available to your organization, an enterprise owner must first enable the feature for the enterprise. See Enabling GitHub Models for an enterprise.
Enabling or restricting models in your organization
You can choose to enable or disable GitHub Models for your organization. You can also choose to only allow the use of selected models or model publishers. For more information, see the instructions below.
You can also integrate your preferred external LLM models by bringing your own keys (BYOK) to GitHub Models. See Using your own API keys in GitHub Models
-
In the upper-right corner of GitHub, click your profile picture, then click Your organizations.
-
Next to the organization, click Settings.
-
In the "Code, planning, and automation" section of the sidebar, click the Models dropdown. Then click Development.
-
Under "Models", in the "Models in your organization" section, click beside Disabled and select Enabled from the dropdown.
Note
If GitHub Models is already enabled for the organization, the dropdown will show Enabled, and you can skip the step above.
-
Under "Models permissions", select one or more options.
- All publishers is the default option and indicates that models from all the current and future publishers available on the GitHub Models catalog from the GitHub Marketplace can be used in the organization.
- Only select models allows you to define a list of models and publishers:
- Available to the organization (Enabled list)
- Restricted for use in the organization (Disabled list)
Depending on your requirements, you can specify an enabled list, a disabled list, or both. Once you've added a publisher to a list, you can fine-tune the list by removing individual models from it.