Yesterday, 12:56 PM
The whole pricing thing for generative tools is crazy because a simple text prompt costs fractions of a cent while generating an image burns way more compute. Going with a flat rate usually ends up with a few heavy users eating all your margins before you even notice. Figuring out the exact billing logic from the start saves a lot of money down the road. You can read a comparison of usage based billings here https://aibilling.dev/ . The site breaks down how different platforms handle token tracking and tier pricing. Just run through the specs there and decide what works best for your architecture.