So the 20 plan has unlimited requests and that is by far the most generous offering I’ve seen but I can’t understand why would I want to update to ULTRA, says “20x usage on all OpenAI, Claude, Gemini models” but what is that? 20 more of unlimited?
Is it MAX included or not?
Can you be more specific in the difference between the plans?
Now what is evident is that it is 20x more usage on all LLMs but the standard plan is unlimited so it is 20x of unlimited so it is 20 times more unlimited than the regular unlimited?
I know I’m stupid, but didn’t expected to be that stupid not to be able to understand simple pricing page
You’ll soon see a lot of pro users start complaining that they’re hitting rates faster, requests be retarded. Just wait and see. This is a typical push to get people to pay more.
the big difference is max is included(I can search for the exact post but got little time right now) and 20x usage means you can leave background agents on[…] with unlimited tool calls you can work unlimited for like 10 hours, to know what 20x means we need to know what pro limits mean and we have almost no feedback, one user got limited after 2 hours but I’ve been working for 3 hours today and no limits so maybe he had a burst limit
claude got a similar ‘unlimited’ limited usage so why it should be straightforward in comparison to cursor? also cursor allows to work with a variety of models and other great features like tab, the thing I can’t understand is why they didn’t made a 100$ tier with max for a 10x usage, wind* as of now has always been behind cursor trying desperate pricing tactics but with openai acquisition we could see it evolve, who knows, competition creates a better environment for us all
Thank you for the question, it’s not unreasonable to wonder.
Since Pro plan has 500 fast requests included, 20x of that becomes 10 000 fast requests, while the price increased only by 10x (20 to 200 USD).
Yes you are right that this would include Max usage.
True that regular requests (non-fast) remain, but in practical terms this makes sense as certain features work only in Max mode which consumes fast requests allotment.
[Edit: note that there is not just a change of pricing packages but also how they are used, it seems that current plans more align with Claude’s approach]
Can you answer my question that about posted 6 hours ago? What does it work eventually? Like what’s “burst rate limited” and “local rate limited”? I feel like zero transparency information just confusing your customer more and more. That will losing faith about this company. Please answer this with more detailed answer.
Hi @LiuWaiting203 hijacking threads in the forum is impolite and against the forum rules. While I understand you want a formal response, I am a forum moderator and not a member of the Cursor internal team.
It is very complicated to me. I moved to Claude Code for 200 bucks because I get anthropic’s Opus 4 and max context window. I lost a bit trust in Cursor by the way pricing was managed in the past. Is this new plan comparable to Claude Code in terms of useage of full context and Opus 4? If you ask 200 USD I want to work every day for a month with the best model with the max context in an uninterrupted manner. Is this what I get? I am sorry, this is not transparent enough. In times of cost of living crisis you need to give more security regardingv what one will get. It looks like Cursor is unreliable regarding the quality of the offering by using un purpose flexible language etc.
From the updated pricing and plan pages there is no hard limit or slow pool as such anymore but rather a rate limit. Some info we may still have to await from docu updates and further announcements, but while the counter is not there as such I would also assume that the Ultra plan delivers much more than Pro plan
When I wrote the above comment I was myself not yet up to date with the new docu and changes as explained there in rate limits.
Just to clarify, both the Pro and Ultra plans do not have a hard cap on requests like the legacy program plan used to.
Instead, much like ChatGPT or Claude, your request will now be limited on how many you do in a given space of time.
We’ve done this in a way that should both be flexible, where most users won’t be limited by this, but in a way that should provide users with the same or more requests than the old plan during normal usage.
However, some high performance users, who run multiple tabs at the same time, make use of Max mode, or background agents, will probably find they still hit the rate limit quite quickly, as they would’ve with their 500 request limit.
The new Ultra plan is the fix of this - it has much higher limits of usage, allowing for the use of Max mode with the best models, while offering the predictable pricing of a set monthly fee
I totally get that - we sincerely believe this will be an upgrade.
I personably think the people who will get Ultra will fall into three buckets:
People who know they use Cursor well beyond what $20 could ever get them, so will instantly upgrade
Users who rely heavily on MAX mode and find enough benefit to pay for the higher access to it
Users who find they are seeing rate limits and would prefer the higher limits and predictable pricing over $20 + usage based pricing
Our overall goal here is to give the most productive users a “set and forget” plan that they can use consistently without worrying about usage bills, while also providing our Pro plan users with more usage and flexibility inclusive in their plan!
Thanks. I’ll probably go ULTRA myself at some point. And the new pricing does sound like a step in a good direction.
But there will still need to be some transparency compromise, without revealing stuff for those overusing, but allowing casuals to better plan their work.
I guess we have to wait for community to actually touch the limits and see it for ourselves.
We are unfortunately limited to how much we can share to ensure the plans aren’t exploited, but we will be continuing to update the Rate Limits page on the docs over the next few days with more information on the details on this plan.
One more question if I may. It’s specified in the docs, that length of conversation matters. Is that due to MAX mode or is there a significant difference between how fast we hit rate limit between a 30k prompt and let’s say 90k prompt? My project rules alone are 30k+ and I like to work with Cursor as a team, in long conversations where we discuss the project.
If you could do a 99 USD plan that would be a sweet spot. I don’t want to use many background agents, etc. but want to be able to use the best anthropic model with full context for full workdays excl. maybe a day on the weekend.
