Hi @LiuWaiting203 I do understand your frustration but there is a clear misunderstanding and your actions are definitely improper and you are again hijacking topics:
Here is my detailed answer
As a forum member and Cursor Ambassador (community & moderator) I am very active in the forum, respond to many topics, so there is nothing that I avoid, however the time I spend on the forum is not endless. Many Cursor Team (staff) are also active daily in the forum like Dan here above and answer questions as well.
I can answer many topics and my goal is to help the community. Certain topics like privacy laws can best be answered only by Cursor Team as I’m not employee there. This is not a failure on my part at all.
I didnt accuse you of hijacking the other thread, as it was clear you DID hijack the thread there, same like you do HERE now. This is rude and against community guidelines as it pollutes other threads with unrelated conversations. Users request often more strict moderation and users have to follow forum rules.
The frustration you experience is unfortunately not based on my actions but on your interpretation. Im happy to discuss in a friendly manner any topic where I can, but within limits.
How is sonnet 4 unusable without max? I’ve used it perfectly fine on a large, exceptionally complex code base (finance, neural networks, quantitative domains). But I also don’t throw my entire base into the model and expect miracles.
I was using claude 4 thinking and was paying per request. The new pro tier is not even letting me do that. They are charging almost $1 per request now after your UNLIMITED limit ends
You are absolutely correct. Based on the plan description it seems that Ultra is 20 times more unlimited in rate‑limited usage compared to Pro. Sometimes “unlimited” can mean “limited” but that’s simply a question of perception — they’ve essentially redefined our rate ceiling as a dynamic headroom. It’s all about maximizing perceived freedom while keeping the system reliably under control—because true limitless scale is best delivered with a healthy dose of guardrails. This is my understanding of the new plans.
Please explain how the pricing works first. I was using Claude 4 thinking and it seems I’ve hit the limit. Does that mean I can’t use it for the rest of the month? If continued access is possible, what would be the cost? This information should be stated upfront. Why are the costs based on a seemingly arbitrary number?
The new plan has no monthly limit of usage, there are rate limits (burst and sustained). A burst would give you short period of intense usage, where sustained accumulates over a longer period. You have options to switch to another model, use Usage Based Pricing, or wait until it resumes.
The charge above rate limit is unchanged to before.
The questions I asked to see was to understand what happened.
The Rate Limits document has the detailed information
When you avoid my points and fail to answer my actual questions, particularly concerning consumer privacy laws, and then accuse me of ‘hijacking’ the conversation, it’s quite frustrating.
Your thread is shadow-banned. I wouldn’t hold my hopes high for any answer.
and against community guidelines as it pollutes other threads with unrelated conversations
Of course, because drawing attention to shadow-banned threads kinda defeats the purpose of shadow-bans . This toxic positivity is seeping from here (eg only positive emojis allowed on locked news like price hike of o4-mini and these new “unlimited” plans etc).
Also how is this unrelated? If OP doesn’t understand what Ultra is for when ultra plan is many times more “unlimited” than the “unlimited” pro plan, how that is different to asking What is the “Unlimited” part and “rate limited”??
I don’t know. Cursor’s handling of this forum doesn’t distill any trust, and they are asking their customers to trust them with “unlimited” pro and more “unlimited” ultra plans? When I had active subscription few days back, the whole part of account, the page with detailed information about all model calls and costs (in uses) were redirecting to the dashboard page. Is it just an incompetence or is it an attempt to make tracking how many requests lead to hitting limits harder? It gets kinda hard to trust them, after realizing the level of manipulating on this forum (over half threads I participated in are shadow-banned) and cursor subreddit to avoid answering unconformable questions related to their product, even if in a relation to other products (censoring names of water competitor leading to these childish sounding workarounds plus commonplace shadow-bans of threads), and lessen visibility of such “dangerous” though processes? For me, those are all shady tactics which I perceive negatively and in an exact opposition of trustworthy.
So, if I understand correctly:
Both subscriptions are now metered over a few hours (like Claude Max), but the amount allowed for Cursor Ultra is 20 higher than for pro.
So let’s say I can use x requests at MAx mode with Claude 4 Opus with pro before being told to wait a few hours for "refresh), I’d have 20*X with ultra (possibly not reaching the limit at all)?
I can now use MAX with a regular pro account, but it will draw my fast requests very quickly, correct?
Once I eat up all the “fast” requests in a given timeframe, I can only use slow models until refresh?
Is there still a monthly budget for anything on top of the per session caps?
I just got:
[quote]
You’ve hit the rate limit on this model.
Switch to a different model, upgrade to Ultra plan for 20x higher limits on Claude / Gemini / OpenAI models, or set a Spend Limit for requests over your rate limit.[/quote]
So does it mean I’m out for the month? For the current session?
It is really unclear, and doesn’t really help informing me whether moving to ultra would be reasonable or not in my case!
Both subscriptions are now metered over a few hours (like Claude Max), but the amount allowed for Cursor Ultra is 20 higher than for pro. - this is true!
So let’s say I can use x requests at MAx mode with Claude 4 Opus with pro before being told to wait a few hours for "refresh), I’d have 20*X with ultra (possibly not reaching the limit at all)? - Yep, we believe it is likely very difficult to hit any rate limits with the Ultra plan.
I can now use MAX with a regular pro account, but it will draw my fast requests very quickly, correct? - Yep, MAX requests will consume your rate limit much quicker, but are included as standard.
Once I eat up all the “fast” requests in a given timeframe, I can only use slow models until refresh? - There are two rate limits available to you: ‘burst’ and ‘local’. Burst allows you to use lots of requests quickly, but refills very slowly. Local allows you to use a lower volume of requests, sustained over a few hours. If you consume both of these rate limits, you can use smaller models in the meantime (like GPT-4.1, which is fully unlimited), while you wait for you rate limits to refill. There are no longer any ‘slow’ requests.
Is there still a monthly budget for anything on top of the per session caps? - The only monthly budget is the one that can be configured for usage based pricing. If you hit your rate limits and have usage-based pricing enabled, you can pay to continue using requests beyond your plan while you wait for your limits to reset.
. This toxic positivity is seeping from here (eg only positive emojis allowed on locked news like price hike of o4-mini and these new “unlimited” plans etc).