So, here’s a new request. Let’s have every model use fast premium requests, each changing based on how much it would cost on the usage based pricing. When the pro subscriber runs out of premium requests, they can then pay more to use the models (either by direct payment like before or by buying request packs) or be limited in which models they can use with the requests slowed down.
Some premium models already charge different amount of premium requests, so why not do it for every model?
Some examples for model pricing if this is implemented:
o1: 10 fast preimum requests per prompt
o3: 7.5 fast premium requests per prompt.
GPT 4.5: 50 fast premium requests per prompt
Sonnet 3.7 MAX and Gemini 2.5 pro MAX: 1.25 requests per prompt and tool call.
And so on.
Similarly, the free requests per day included in pro plan for some models can also be removed.(I doubt anyone is using those models anyways)
I believe this would help reduce the confusion about Cursor’s current pricing model.