LLM Models
Manage which models this organization can use.
Accessing the Page
Admin → AI&MCP Settings → LLM Models
Configuring Model Access
The models available to your organization are displayed in a table. Use the toggle switch on the right side of each model to enable or disable it.
- Enabled: Organization members can select and use this model in chat.
- Disabled: This model is removed from the model selection list in chat.
Enabling Auto Router
Auto Router is a model that automatically selects a model based on the user’s request and conversation context. It is disabled by default. To let organization members use Auto Router in chat, an administrator must enable it first.
To enable Auto Router, turn on the Auto Router toggle in the model list, then click Save Changes in the top-right corner. After the change is saved, Auto Router appears in the chat model selector for organization members.
The models used for each Auto Router tier are currently set by the system and cannot be changed from this page.
Regional AI Processing
In addition to general chat, where users select a model directly, internal AI tasks are performed for service operation, including generating conversation titles, reviewing conversation context and memory, and selecting tools for requests.
These internal tasks may use models selected internally by the service, such as gpt-5.4-mini and text-embedding-3-small.
When Regional AI Processing is enabled, supported internal AI tasks are processed using Japan-region models. This setting is separate from the model users select in chat and is intended for organizations that need internal AI processing to take place in the Japan region as well.
Turn on the Regional AI Processing toggle, then click Save Changes in the upper-right corner.
Note: This setting does not change the model that users select directly in chat. Image generation and editing currently use the existing image generation model.
Behavior When a Model Is Disabled
- Chat: If a model previously selected by a user is disabled, they are automatically switched to the default model.
- Agent: Only models permitted by the organization are shown in agent settings. If a model assigned to an agent is disabled, execution is blocked and an error is returned at runtime.
Saving Changes
After making changes, click the Save Changes button in the top-right corner to apply them.
Note: If you attempt to leave the page with unsaved changes, a warning will appear to prevent accidental data loss.
Summary
- Enable or disable individual LLM models per organization using toggle switches
- Disabled models are immediately removed from the chat model selection list
- Changes must be saved by clicking Save Changes to take effect