Refunds and billing

What you are charged

Exactly twice what the model provider charges us, worked out from the usage the provider reports for your request. Never an estimate applied after the fact. If a provider reports no usable cost, you are charged nothing rather than being handed a number we made up.

Generations quote first

A picture or a video costs tens of tokens and takes minutes, and once it starts the provider bills us whatever you make of the result. So it is quoted before it runs, and nothing starts until you agree to the number.

When a model fails

• It never started. Nothing is charged — you cannot be billed for a request that did not reach a provider.
• It failed midway. You are charged what the provider actually billed us for the part that ran, doubled. Usually that is a fraction of the quote.
• It finished but the result is bad. That is charged. A model that answers badly has still done the work, and we cannot recover it from the provider. If a model is consistently wasting your balance, tell us — that is a catalogue problem, not a billing one.

The free assistant

Talking to AiHub AI costs nothing, ever, including when it is being served by a paid model behind the scenes. That cost is ours.

Something looks wrong

Your ledger shows every charge with the provider’s own cost underneath it, so the doubling is visible rather than promised. If a row looks wrong, send us the date and the model at billing@aihub.onesmm.ai and we will look at the actual provider record behind it — though the domain has no mail route today, so that address does not deliver yet. The contact page says so about every address on this site.