Model Fallback

Learn how Model Fallback automatically switches to a backup AI model during provider outages to keep your allmates.ai conversations running smoothly.

Last updated About 3 hours ago

Occasionally, AI providers experience temporary overloads, timeouts, or outages. Instead of interrupting your workflow with a red error message, allmates.ai uses Model Fallback to automatically switch to a backup model. This ensures your Mates can always reply to your prompts without delay, providing a seamless and reliable collaboration experience.

How It Works in the Chat

For standard users, the Model Fallback process is completely invisible and requires no manual action. If your Mate's primary model is unavailable, the conversation simply continues using the backup model.

To maintain total transparency, you can easily spot when a fallback has occurred:

  • Look for a small amber icon in the message header, located next to the token counter.

  • Hover over the icon to reveal a tooltip. This will tell you exactly which backup model generated the response and which original provider was unavailable.

Configuring Model Fallback (Admins)

By default, Model Fallback is enabled and set to "Automatic" for all organizations. If you are an Organization Owner or Admin, you can customize this behavior to fit your team's specific needs.

  1. Open your Organization Settings.

  2. Navigate to Models Management, then click on the Settings tab.

  3. Locate the Fallback Model dropdown (positioned just below the Default Model setting).

  4. Select your preferred fallback behavior from the available options.

Available Options

Automatic

(Default) allmates.ai automatically selects the best available backup model to ensure continuity.

Specific Model

You manually choose a specific model from your available list to always act as the backup. (Note: The current Default Model is excluded from this list to prevent loops).

None

Disables the fallback feature entirely. If a model fails, the standard error message will be displayed to the user, requiring them to retry manually.

Important Details

  • Scope: Currently, Model Fallback applies only to standard conversations with Native Mates. It does not yet cover specialized features like Image Generation or Web Search.

  • Single-Level Fallback: There is only one level of backup. In the rare event that both the primary model and the fallback model fail, the standard error message will be displayed.

💡 Tip: If your organization relies on a specific tier of models for data privacy or compliance reasons, use the Specific Model option to ensure the fallback always aligns with your security policies.

Related Articles

  • Managing Organization Roles and Permissions

  • Understanding Mate Knowledge and Context