Quality checks

A useful tool earns its place through actual performance.

What we look at

Supakook focuses on account availability, request performance, and model answer quality. We take a zero-trust approach to upstream resellers and use official accounts we hold ourselves, reducing reliance on supply chains of unknown origin.

Scheduled checks

What is checked	Schedule	What we observe
Codex	Hourly	Changes in answer performance using the fixed “pelican question” test prompt
Web Pro models	Daily at 08:00	Actual answer performance in the web interface

This is the check schedule provided by the service. This page does not display live monitoring results. Refer to the service delivery instructions or published check records for the time zone used for the 08:00 check and actual run records.

How to interpret results

A fixed test prompt can help reveal changes in performance, but a single test cannot represent every capability. Differences in answers may reflect model versions, prompts, context, tool settings, or randomness. One response is not enough to judge a model's overall quality.

A model naming itself in a conversation does not prove its source or version. Account origin, actual routing, and answer quality need to be checked separately; matching request fields alone cannot establish them.

If you notice an issue

Record the time, product, model identifier, and steps to reproduce the issue. If possible, include a comparison prompt containing no sensitive information, and report it through your original order channel. Do not submit complete keys or business materials that still contain sensitive information.

Whether an issue is covered, and how it is handled, depends on the service coverage information and the agreement supplied with your order.