Spend limits

This page publishes the automatic spend limits that Shara applies. It is the address the subscription contracts (First, Pro and Max) and the Concerto Local contract refer to, which is why they no longer carry the figures inside: that way the paper cannot contradict the system the day a rate changes. The contracts cite the Spanish version of this page; this translation shows the same figures, read from the same source.

The figures are expressed in STU, the unit Shara uses to measure model consumption. The limits are there for protection, not as part of a plan: they exist to stop an error or a loop from eating anyone’s budget.

1. The limits in force

2. How to read them

• All three are evaluated at once: exceeding any one is enough for it to act.
• They act in the Customer’s interest: they cut spending, not the service contracted.
• These are the default values. The owner or an administrator can change or disable each one in Settings.
• The Customer can ask at any time for all inference on its account to be suspended immediately, without waiting for any limit.
• In Concerto Local they only apply to tasks that run in the cloud. Work done by the Customer’s local engine is not subject to them, because it generates no inference cost for AIGiner.

3. Where this figure comes from

From the same constant the engine applies the cut with, not from a copy written by hand. When that constant changes, this page changes with it, and an automatic check prevents the website from being published if the two stopped matching.

If at any point what you read here did not match what the Service does, write to us at support@aiginer.com: it would be our mistake and we treat it as such.