Hi, very new user here. I very quickly burned through my free 150 premium model used by using the agent tool. Now, when I started, I didn’t realize what the usage even was or how this all worked, so I certainly could have used the ask tool and some of the free models for many tasks.
I am more than willing to upgrade to Pro but am confused on how it actually works. It seems like I will get 500 premium fast usage per month… but once the 500 is used, can I still use the agent tool with some models but have slower responses? Is that the differentiator between pro vs hobby? I want to continue using the agent tool and am more than willing to use slow responses. The only tools I use in Cursor IDE are agent & ask. The ask tool, using cursor-small was not performing very well and I ended up just using chat gpt 4o for my tasks (on the chat gpt website)
Just want to understand and clarify before I commit to the Pro plan. Should I be using some specific model to reduce the amount of premium requests I am using?
Yes, after those 500 you can still use the same models, and depending on time of day / number of requests you did, they will be slower or faster.
Honestly, using almost exclusively Gemini 2.5 and agent mode for the last whole month (over european daily hours), I didn’t really notice the difference between fast and slow requests that much.
Firstly, because with a thinking model, you will probably wait more through thinking process than any slowdown due to potential queue.
Secondly, because having to wait for AI response does’t mean you can’t do anything useful in the meanwhile.
Should I be using some specific model to reduce the amount of premium requests I am using?
Focus on your prompt-fu, project rules (including understanding of your project), and invent some simple task management system for AI to follow, based on your needs. Make the models actually use those requests well. It doesn’t have to be to their limit, just make them reasonably do a lot rather than just a little. And you won’t have to use specific model just to reduce the amount of requests.
Remember you can stack orders for AI inside one request. It doesn’t have to be “Please do this.”
It can be “Please do this, this, and this, and later that, that and that. And once you’re done, go through all of our remaining tasks for this feature.”, and it will count as one.
Thank you for this detailed response. This is very helpful!
1 Like
Are you using Gemini 2.5 flash preview, pro 3-25, or pro-max?
Pro 03-25, which now is 05-06 under the hood. But currently after the update it’s really playing on my nerves, and others have the same issue with the AI model unable to properly use file searching/writing tools and stopping in the middle of a response.
Sad Panda.
1 Like