I’m seeing a lot of bad reviews about pro being rate limited within 2-3 messages. Is it working properly for anybody else. I wish Cursor team would clarify what the rate limits are. From what I understand:
$20 worth of request (makes sense)
+
local limits (4h) (how much?)
+
burst limits (24h) (how much?)
which makes sense, but why are you not telling us exactly how many requests are allowed in these rate limits. I am currently a free user, and don’t want to use pro if its going to give me like 2 requests per day. Also, I know more expensive models are lower rate limits, so please tell me rate limits for:
Claude 4 sonnet - normal
Claude 4 sonnet - MAX mode
Claude 4 opus - MAX mode
Gemini 2.5 Pro - normal
Gemini 2.5 Pro - MAX mode
R1 0528 - normal
o3 - normal
o3 - MAX mode
o4-mini - normal
o4-mini - MAX mode
gpt-4.1 - normal
gpt-4.1 - MAX mode
“Auto” model - normal
(Sorry for requesting pricing for so many models but I’m sure you have a table of it somewhere, and even if you don’t, I should have a right as a potential paying customer to question what I’m paying for, especially with competitors who have a much clearer pricing scheme)
I think it’s better at this point to stick with a single model. Get the API key for that model and switch to using Roo-code instead. The current Pro plan doesn’t seem to offer much value, and their pricing calculations feel unclear. Additionally, it looks like Cursor has started using some questionable dark patterns.
Exactly My friend. No one wants to get rate limit after 2-3 messages. You can check my posts on my profile. I’m opening topics because of that conversation you talking about. But I didn’t get – No one didn’t get – any answer. So No one will prefer to buy Pro plan… People want to get the value they deserve.
Ok, I know but I want to see if there are any good reviews about it since everyone is saying it’s so bad. Wanted to check with Cursor themselves. And I think W!ndsurf and Trae is better because their pricing is so much clearer now. (oh, btw Roo Code might turn out more expensive than some of these IDEs)
but, I’ll still hang around incase something good happens
I’m going back to the old billing model or canceling. I can’t work with the new model anymore. I send 2 or 3 messages and I already hit the rate limit using Sonnet 4 (not at MAX).
Nah, I wouldn’t say that this is the concensus. There are some users who’s usage falls in between plans, and thats not unexpected given Cursors large userbase.
Some issues reported by users were fixed and for others that requested a plan between Pro and Ultra Cursor has added Pro+, but that depends really on the usage need.
Overall the new plan works for most better than with the previous slow queue.
I’m using my personal Pro plan and have never hit rate limits (posted in various threads about my usage) with over 15000 lines edited by Agent per day. Actually this month I wont have any usage based pricing addon (usually 20-50 USD). While I am the Ambassador here in the community, I have no insight into Cursor internals, so I dont know the limits either.
I’m no longer able to make progress with Sonnet 4. I literally send 2 or 3 messages and get blocked by the rate limit (even waiting 24 hours between interactions). I thought it was an issue with my MCP servers that I’m using, but that’s not the case.
At the beginning of the switch to the new billing model, everything was absolutely fine. But not anymore.
While we’ve not yet posted the structure of our rate limiting system, I can explain how the models work!
The impact a request has on your rate limits is almost exactly corrolated to how much it would cost via an LLM provider’s API:
How long the specific message is
How long the past conversation is
How many files are attached
How much does the specific model cost (thinking = higher cost)
Is MAX mode (longer context cap) enabled?
While this doesn’t match or use our old pricing system, that can be a good indicator of the price of a model!
Claude 4 Sonnet Thinking is quite an expensive model, and has a decent context window, which can make it a higher-cost option - previously this used 2 requests in the old plan.
o3 is a slower but still quite intelligent model, but has a lower usage cost - previously 1 request - so will have a lower impact on your usage.
GPT-4.1 and Gemini Flash are both totally unlimited too, and don’t touch your rate limit at all.
Finally, I’d highly recommend people use Auto, as this always provides a “premium” model (Claude 4, Gemini 2.5 Pro, etc) but at a significantly reduced effect on your rate limits - this is because we can intelligently route to models with lower usage and utilisation!
A note for those who are finding they can only do a few requests before being rate limited, this is a pattern we have seen a few people fall into - let me explain what may be happening!
As stated in our documentation, your usage is made up of two rate limits:
Burst Limit - A very large, but slow to refill limit (refills over days)
Local Limit - A smaller but faster refilling limit (refills in a few hours)
For high-volume users in the first few days of the plan, especially when testing the limits of the new plan, many have chosen the more expensive models like Claude 4 Thinking, and in turn, have burnt through their entire Burst limit.
This unfortunately then leaves you with your local limit, which when using the same expensive models with long conversations, will also be used up quickly.
Where possible, I’d recommend avoiding the temptation to use expensive models for simple tasks (especially thinking and MAX mode). Alternatively, using Auto as described in my post above can have a significantly reduced impact on your rate limit while still providing an intelligent, premium model to respond to your request!
While I unfortunately have no further info to add, I think this is super important to highlight, as this can really make a huge difference to your rate limits while still getting premium models every time:
Maybe before changing pricing the client should have had an update to reflect this pricing?
The pricing still shows this request based pricing with o3 and sonnet-4 being the same price. From what youre saying here its not the case.
Some of my requests now are 20x the price as they were previously.
Auto is hopeless and gets things wrong 90+% of the time.
This is on top of not usually using the thinking models and the IDE changing me over to thinking automatically for Sonnet-4. I didn’t even realise for days because I had never seen the brain icon before.
I had to go into my settings to reactivate standard sonnet. So this was forced on me.
The latest changes have been super unclear with old information lingering, have forced people into using more expensive models and costing much more in the process.
Question about the legacy mode, is this 100% what it was before (500 requests + $.04 per request after) or is it on the 500 requests and still paying upwards of 20x per prompt?
Ive just changed over to the old plan and I have 273/500 requests used with $45 usage based cost and only 10 days left this month.
This shows how much worse the new payment is.
Heres some context for what I’ve been on this month, let me know if this gives you a clear picture
Current rate limits + $45 use base cost.
Memory bank which has a work sequence of Plan → Creative → Implement.
When I get around to Implement it’s not unsusualy for me to use multiple 25x tool calls.
I can opt back into the new pricing the same way I opted out.
I don’t know about the pricing after reaching 500 requests unfortunately..
I doubt it will go back to $0.04 per request though, I have asked but I never get replies from mods or devs
Hopefully this gives you enough context.
Old pricing is better if you stick within the 500, I might have to optimise my prompts some more and hope it stays around…
As stated in our documentation, your usage is made up of two rate limits:
Burst Limit - A very large, but slow to refill limit (refills over days)
Local Limit - A smaller but faster refilling limit (refills in a few hours)
This is annoyingly unclear. I can’t tell if you folks are delusional or if this is an honest UX/product mistake (in which case, you’d be working to solve it ASAP!). How many days? How many hours? If customers want to upgrade, they need to know this, otherwise, why am I going to upgrade if I don’t even know the limits of the higher tier?
The lack of transparency is quite concerning here lol. What was even the point of changing the plans and rates? Why break everything when it was working fine. I must be getting old cause nothing makes sense anymore.
