Available Models

Workspace owners and admins decide which AI models members can use. The Available models control sets the list of models members can pick in the model picker for chats and tasks.

Find the setting

Open Settings → Workspace → AI Controls → Available models.

The page lists every model, grouped by family (Anthropic, OpenAI, GLM, and others).

Note: Only workspace owners and admins see this page. If you don't see Available models in your settings sidebar, reach out to help@obvious.ai to see if it can be enabled for your workspace.

Set the available models

Two controls decide what members can pick.

The model switches turn on the models members can choose. Members can't pick a model that's switched off.

Anthropic, OpenAI, and GLM also each have an Auto switch. The Auto switch automatically picks a model within that family, and it can use any model in that family, including models you've switched off. That's why it has its own switch: you can turn off a family's Auto option without changing which individual models members can pick.

Both settings are optional: if you set nothing, members can pick every model in the catalog. You can't save a list with zero models. At least one model has to stay available.

Changes save as you make them. If a change can't be saved, a message appears and you can try again.

Tip: Use Enable all to clear all limits and make every model available again.

Models that are always available

Fusion and Fusion Pro are always available to every member. They can't be hidden or restricted.

What members see

• The model picker shows only the models and Auto options you've left available.

• Members can set a default model for new chats and tasks in Settings → Behavior. The list they choose from respects your Available models. Admins set the limits; members set their default within them.

• New chats and tasks start on the member's default model when the member hasn't picked one.

• If a chat or task was using a model you later removed, it runs on the member's default model, or the first available model, the next time it runs. No error appears. Runs already in progress keep the model they started with.

Removing a model never breaks an existing chat.