Skip to main content

​
Our Promise: You Only Pay for Value

In the world of AI, not every API call yields a perfect result. Models can fail, requests can error out, and sometimes, responses are just empty. We believe you shouldn’t have to pay for these outcomes. Our Zero Completion Insurance is our commitment to fair and transparent billing. It’s an automatic safety net that protects your budget by ensuring you are only ever charged for successful, valuable responses. This feature is enabled for all accounts by default. No configuration is needed.

​
The Simple Rules of Protection

Our insurance policy kicks in automatically under two simple conditions. We will never charge you for a request if:
  1. The response is empty: The model returns zero completion tokens and the finish_reason is blank, null, or indicates an error (like length for exceeding token limits).
  2. The response is an error: The upstream provider returns an HTTP error code (e.g., 4xx or 5xx), such as a rate limit exceeded, a server failure, or an invalid request.
This applies across all models from all providers, giving you consistent and predictable billing.

​
Why This Matters

For developers building real-world applications, this insurance is a game-changer:
  • Experiment with Confidence: Freely test new models and complex prompts without the fear of racking up costs from failed attempts.
  • Handle User Input Gracefully: Process unpredictable user-generated content without worrying about paying for invalid inputs that cause the model to fail.
  • Scale-Up Safely: Run large-scale, high-concurrency jobs knowing that you are protected from intermittent provider failures and errors.

​
Complete Transparency

On your account’s activity page, any request covered by this policy will clearly show zero credits deducted. This is true even in cases where the underlying provider might have charged us for processing the prompt. We absorb that cost for you.
⌘I