# Model availability

> Decide which AI models your workspace, a team or one person can use. Media and chat models are separate catalogs, and each scope can only narrow access.

Source: https://endless.okinawa/docs/governance/model-availability

**Settings → Governance → Models** controls which models the workspace, a team or one person can
use, and which model is the default. Admins use it to control cost and to keep everyone on the
same models.

<Callout type="info">
  Opening the page needs **View workspace activity and governance**; changing it
  needs **Change workspace governance settings**. Without the second, the page
  is read-only. See [Roles and permissions](/docs/governance/roles-and-permissions).
</Callout>

## Turn a model off

1. Pick a **Scope**: **Workspace** (the default for everyone), a **Team**, or a **Person**.
2. Open the &#x2A;*"Media models"*&#x2A; or &#x2A;*"Chat models"** tab.
3. Switch the model off.

To allow only a few models, use the switch at the top of the media tab to turn every media model
off, then turn on the ones you want. The media tab is grouped by kind (image, video, audio and so
on), each section with an &#x2A;*"x of y"** count of enabled models.

Media models show their cost per run in credits, or &#x2A;*"Varies"*&#x2A; when the price depends on the
input. Chat models show a relative cost in &#x2A;*$** signs: more $ means more credits per message.
Current prices are on the page; see [Models](/docs/concepts/models) for how to choose.

## Choosing the default model

Drag models to reorder them. The order is the order of the pickers, and the first enabled model of
each type is marked &#x2A;*"Default"** and is used when nobody picks a model. On the **Chat models**
tab, the first enabled model is what new chats start on.

* A team or person uses the order of the scope above it until someone reorders at that scope.
  From then on, that scope's order replaces the inherited one.
* A model that needs a plan the workspace doesn't have is skipped, so the default is always a
  model the person can run.

## What turning a model off does

* The model disappears from every model picker and from `@` mentions, and chat can't use it.
* Requests that name it are refused by the server, so nothing runs and nothing is charged.
* Work already made with it isn't affected.

## Rules

* **Restrictions only subtract.*&#x2A; A model turned off for the workspace stays off for every team
  and person. The row says which level did it: &#x2A;*"Off for the whole workspace."*&#x2A; or &#x2A;*"Off for
  this person's team."** To give one team a model, turn it on for the workspace and off for the
  other teams.
* A row marked &#x2A;*"Inherited"*&#x2A; follows the scope above. &#x2A;*"Off for this scope"*&#x2A; means the switch
  at the scope you're viewing turned it off. &#x2A;*"Reset"** clears both the switches and the order
  set at that scope.
* Members can turn models off for themselves and set their own order on the **Features** tab of
  their own profile in **Settings → Members**. They can't turn on a model the workspace, their team
  or an admin turned off.
* **Turned off here vs. needs a plan:** a model that needs a paid plan stays in the picker, dimmed,
  with the plan's name. A model turned off on this page isn't in the picker at all.
* To stop a whole kind of work, such as video, turn the tool off in [Tool
  availability](/docs/governance/feature-flags) instead.

## Related

* [Models](/docs/concepts/models)
* [Tool availability](/docs/governance/feature-flags)
* [Teams and limits](/docs/governance/teams-and-limits)