Skip to main content
Zero Completion Insurance is a default protection that ensures you are never charged for requests where the model fails to produce any output due to a provider-side error.

​
How It Works

If a response meets all of the following criteria, the request is not charged:
  1. usage.completion_tokens == 0 (the model produced no output tokens)
  2. finish_reason is empty, null, or "error"
  3. The failure originated on the provider side (not from an invalid request)
In these cases, usage.cost is returned as 0 and no credits are deducted from your account.

​
What’s Covered

ScenarioCovered?
Provider timeout with no outputYes
Provider internal error (5xx)Yes
Model refused to respond (safety filter, 0 tokens)Yes
max_tokens: 1 with 0 completion tokens returnedNo (valid request)
Your request had an error (invalid model, bad JSON)No (client error)

​
Viewing Protected Requests

In the Activity page, requests protected by Zero Completion Insurance show:
  • cost: $0.00
  • finish_reason: error or finish_reason: null
  • completion_tokens: 0
You can filter by cost = 0 and completion_tokens = 0 to audit these requests.

​
No Configuration Required

Zero Completion Insurance is always active. There is no parameter to enable or disable it. If you believe a request should have been covered but wasn’t, contact support@arouter.ai with the generation ID from your Activity log.