Even though I am on the Pro plan and the model panel shows “unlimited” for Claude 4 Sonnet, I consistently receive a “You’ve hit the rate limit on this model” message after only two prompt submissions. This happens repeatedly and disrupts usage.
Steps to Reproduce
Open Cursor with a Pro subscription
Select the Claude 4 Sonnet model from the chat dropdown
Enter a prompt and submit
Submit a second prompt
On the third prompt attempt, the rate limit error appears
Expected Behavior
As a Pro user, I expect to be able to use Claude 4 Sonnet without hitting a rate limit, especially since it’s shown as “unlimited” in the UI.
Operating System
Linux
Current Cursor Version (Menu → About Cursor → Copy)
Looks like you used far more than the equivalent limit in the old plan with claude-4-sonnet-thinking model usage between June 18th and June 25th, which would have used up your burst rate limit:
Docs on this here:
You’ll still continue to get local rate limit refreshes every few hours.
Thank you for posting this. I casually use cursor and hit the rate limit after a few minutes of simple requests. I usually would spend the pro $20 and maybe another 5-10 on usage to avoid the slow queue per month. Now my usage based pricing already hit a few dollars in a few prompts. I use cursor maybe an hour or two a day here and there as I work on a game project.
At this rate id spend WAY more for the same casual use or i have to work for 5 minutes and come back in a few hours. I cant justify this for a hobby.
Hey, checking your usage, I can see you’ve used over 300 requests in the last ~2 days which has likely drained your “burst” rate limit allowance, which is a large but slow refilling limit you have that is intended for quick, high volume usage.
At that level of usage, you would likely consume your 500 requests in under 7 days, suggesting you’ve either had a very heavy few days of usage so far, or you’d be spending >$40 a month of usage-based pricing to sustain this.
If you only use Cursor in spikes throughout a month, then you should find that your burst rate limit will recover in the near future, but your current usage sustained throughout the month would be way above the limits of the base Pro plan, both in the old and new pricing systems.
If you have any insights on your usage here, I’d be happy to explain further how this may work for your specific situation!
I don’t think I need to mention that it’s terrible UX (and I’d say business practice) to just sell Pro as unlimited, hit the user with a rate limit without any information of when the limit refreshes (like Claude Web, ChatGPT web both do).
This makes the Pro plan’s “unlimited” claim feel misleading, since users can’t even anticipate or manage around the rate limits.
While the exact values aren’t available, the “burst” limit refills very slowly but is large in size, whereas the “local” rate limit is smaller, but refreshes every few hours.
The goal with this change is to hopefully allow users to not have to manage their requests as much - if a user is finding themselves hitting rate limits often, they likely get more than $20 benefit our of Cursor, and therefore have the option to upgrade to Pro+ or Ultra to enable to continue using Cursor at the same pace.
To be clear, this new system allows users to get >500 requests within a month during normal usage!
Users might upgrade out of desperation. Other users might just abandon the product. Even if you upgrade to the $60 or $200 plan, how do I know I’m getting $60 or $200 worth of usage if I don’t even have a reference point (of when the limit refreshes, for example). Very confusing and shady.
It’d be much better if the user could see the refresh time in each of the plans.
And don’t get me wrong, I enjoy Cursor very much, and I even think the plan changes wouldn’t be bad if you were more transparent on the actual limits. It seems you’re focusing on “unlimited but with throttling”, so if you really want to sell it as unlimited in this context, you should let users know how it works, when it refreshes, otherwise, it’s not unlimited.
Unlimited doesn’t mean unreestricted. That’s how it has always worked with nearly everything. Your phone carrier and home isp both have limits no matter how fast the data is or how unlimited they call it.
17 requests per day, is normal usage if you average it out over the month.
It’s clearly different - with my carrier I know the limits and I know what is the speed it will fall into when the limits are reached (in their “unlimited” plan). With the current plans in Cursor, we don’t know when the limits are reached, exactly, and we don’t know when they are restored.
And by the way, my home ISP doesn’t impose any limits whatsoever. I know it’s a thing in the US, but still not common practice in most other countries
I’ve noticed that on average with the current limits I can perform 15 queries to Sonnet 4 per day, which is 450 over 30 days, so… can anyone confirm that Cursor’s competition still offers 500 AI generations for $15 and they have Sonnet 4 in their offer?
I really want to ask, I’m a pro user using sonnet 4 most of the time, and after the new plan comes out, the spending seems to become much higher. Same amount of usage compare to last month, but last month I spent about 20 dollars, and this month it already costs me around 100$, and I see some sonnet requests costs more than one dollar or even more than two dollars.
If allowed, I would choose to quit, but I purchased 1 year of use and there is no way to quit. I have used more than $50 in the first 10 days of this month.
