​
How It Works

  1. Your primary model gets the request.
  2. If it fails (timeout, error, etc.), the router immediately tries the next model.
  3. This continues until one model delivers the results you need.

​
How Does This Help?

  • No more stalled workflows
  • Higher success rates
  • Optimized costs - you only pay for what works

​
Get Started

  1. Go to Manage API
  2. Add a Fallback Policy (see screenshot)

Here’s an example setup:

  • Policy Name:

    Deepseek-R1

  • Fallback Chain:

    deepseek/reasoner → nebius/DeepSeek-R1 → deepinfra/DeepSeek-R1

Make sure each model in your Fallback Chain is compatible with your request parameters (e.g., context length). If a particular model cannot handle the request’s requirements, the policy will move on to the next model without additional warnings.

Previous
Load Balancing
Next