Have you heard about “Tokenmaxxing”?

Since many AI companies didn’t have a reasonable limit on the number of tokens for the amount of money you paid, companies started telling their employees to use as many tokens as possible. LLMs improve with more tokens (although there are diminishing returns).

So users tried to exploit the ‘lure offer’, and AI companies had to change the billing. It’s still below the real cost, but no longer this insanely expensive option.