Yes, pre-paid would be fine and it's a well-understood pattern.

No need to retire API keys.

Implementing this in any meaningful manner quickly begins to look like every read becoming a globally synchronised write. Of course it doesn't have to be perfect, but even approximating perfection doesn't look much different. Also, can you imagine the kind of downtimes and complaints that would inevitably originate from a fully synchronous billing architecture?

Open ai has this

Prepaid only is a fantastic idea, until your site goes (desirably) viral and then gets shut off right as traffic is picking up, or you grow steadily and forget to increase your deposit amount and suddenly production is down. Billing alerts are a much better solution IMHO.