You will always be charged your based plan amount of $20, regardless of how many requests you use. However, instead of giving you the set 500 requests a month, there is no longer a hard cap per month, but much like ChatGPT and Claude, we now enforce rate limits for how many requests you can do a set timeframe.
The only time you will ever need usage-based pricing is if you attempt to use more requests than you can with the rate limiting of your plan. In this case, usage-based pricing can be enabled to fill the gap between the $20 and $200 a month plan.
To summarise:
- The new Pro plan allows for more included usage than the old one, spread in time across the month
- If you hit your rate limits, you will be asked to switch models or enable usage-based pricing
- The majority of users should not be hit with rate limits often