Clarifying June 16 Pro Changes

Discussions  Featured Discussions
1

Hello! On June 16, we rolled out a few changes to our Pro plan:

  • The plan moved from request limits to compute limits. We worked with the model providers to give all users at least $20 of agent model inference at API prices per month. You will never see a rate limit when you’ve used under $20 of inference.
  • We lifted all limits on the number of tool calls per agent request.
  • We introduced unlimited access to our ‘Auto’ model switcher.

Our communication around these changes created confusion, and we take full responsibility for this.

Today, we’ve updated our docs / website in a few ways to improve our clarity and better set expectations (pricing page, pricing docs, model docs, ultra blog).

12 Likes
2

Thanks for the increased clarification.
One question, the docs state:

There are three individual plans: Pro ($20 / mo), Pro+ ($60 / mo, 3x limits), and Ultra ($200 / mo, 20x limits).

But there’s no mention of Pro+ on the pricing page. Is this plan new?

2 Likes
3

Thanks for updates! Allow me to ask some questions.

For example, Pro includes over $20 of agent model inference at API prices per month

What does it mean? Can it be written with simpler wording?

We lifted all limits on the number of tool calls per agent request.

So there’s no longer internal limit of 200 tool calls?

All individual plans include: Access to Background Agents

(…) few sections in between

Background Agents are charged at API pricing and aren’t included in any plan for the time being.

What does it mean? If it only means “All invidual plans include ability to PAY extra for Background Agents”, can it be included at the top?

Max Mode is included in all plans. Turning it on will pay down your agent limits slightly faster, since the API cost of your requests will go up a bit. The difference should only be noticeable for 1M token context window models.

What does it mean? Provided only Gemini and GPT 4.1 have access to 1M context window.

Does it mean that for other models like Claude 4 Sonnet there should be no noticeable difference when using MAX mode?

Does it mean no noticeable difference between two requests (1 MAX, 1 non-MAX) on the same prompt length? If yes, why is it different for Gemini and GPT 4.1?

Does it mean no noticeable difference between two requests on different prompt lengths, unless they reach close to 1M? (f.e. 1 at 50k context length and 1 at 200k context length)

Thanks a lot!

7 Likes
4

This doesn’t help, sorry.

5 Likes
5

The new pricing system is extremely obscure. I know this is not what the Cursor team intended, but from the perspective of an active customer, it feels like a bit of a shell game. The fact that I had to read through four different documents just to understand what’s going on really illustrates the problem — don’t you think?

I shared my experience in this forum post. I love Cursor and really appreciate the work you’re doing, and I sincerely hope you’ll take the time to do a proper review and fix these mistakes.

8 Likes
6

One more question.

Explicit error messages are always shown when users hit rate limits.

I’m on old plan, so I don’t hit any rate limits but… isn’t that vastly different from what many users experience, and what was already confirmed by the team, that there is no message when usage based charges are on?

Did something change, or is this sentence inaccurate?

1 Like
7

I’m curious about how yearly payments work under the old Pro plan. If someone paid for a full year and now switches to the new pricing structure, what happens to the remaining balance? Does the “budget” expire?

1 Like
8

Hey Team,

Thanks for thee update!

However, I’m puzzled with the outcome. Or perhaps we don’t have the same definition of what “clarification” means ? :thinking:

There is still no clear explanation (i.e: given numbers) of the newly introduced rate limits (How much, and how long it takes to refill).

What means X3 and X20 if we don’t know what to base these multiples on?

Why don’t you implement something as simple as the GPT model (You’ve exhausted your “X” limit and it will resume at X HH:MM) ?

Currently, you’re asking us to read the pricing plan in a dark room, with no light, no interruptor and holding the lighter in a different building.

Alex

3 Likes
10

Since the cursor team has to bear all the consequences and responsibilities caused by the confusing pricing explanation, should the extra money we are charged be returned, or compensated to consumers by increasing pro membership time? If it is just a simple update of some documents, this cannot be a reasonable explanation for taking all the responsibilities.

2 Likes
11

Yeah this really isnt the context we were looking for. It’s still very vague.

Whats the chances that a $40 Pro(Kinda+) will be introduced? for those of use that could do with more but dont want to pay $60 a month?

One of the best things about the previous models (other than must cheaper requests) was that if I didn’t use Cursor much 1 month I only paid 20 instead of 60 when I max out my limit.

1 Like
12

@bengin And not to mention the bug which made it so you couldnt see your cost for this month, which im sure I’m not the only one who spent more money because of.

13

+1 hehe

14

Hi Reznal
I have now switched to the old pricing model. Although it still costs a lot of money, it is much less than the new pricing model. I didn’t realize that the so-called unlimited request number was just empty talk until I was deducted 40-60 US dollars by the new pricing model. I have been communicating with the cursor team and asked for a refund of the extra money, but so far there has been no reasonable explanation or follow-up.

15

It’s not clear, if Pro+ or Ultra actually have Opus 4 access (without usage based pricing), and context window.

That’s what most users are looking for after Claude Code stepped up. People want same quality with a GUI.

Good luck w Anthropic!

2 Likes
16

I used to get 21 requests a day 500/30 ish. Now I get MAYBE 3 a day waiting on the queue. I understand why and what you’re trying to do, i just don’t think that makes it a viable plan. Also, I’m completely unable to find this alleged “Auto” mode. it doesnt give me the option lol.

1 Like
17

It didn’t help at all lol, we still don’t know how much we have left, nor the real size of the tokens we are entitled to, nor how long it will take to recharge them.

1 Like
18

One question: how long does it take for the burst limit to refill?

Hiding this makes zero sense, as we will find out as soon as the first person’s resets.

Big thank you for getting back to us!

3 Likes
19

All I see is blanket terms and “patches” of surface level conversations that seem to be more deflecting than actually being productive on the matter. I understand fixes and changes are being made and that technology still hasn’t hit its peak in terms of LLM architecture and “layers” of AI memory able to calcuate at “x” speeds utilizing “x” cost per api etc. and everyone has to make money to continue development.

But if I pay $ and this can go from $0 to $1000 based on the model used, I would like to be more “conservative” but when the only thing I know is I’m “burning” more and more money, how am I suppose to be “satisfied” in this regard?

The transparency on the supposed “Limits” seems to affect not only newbies as myself but even veteran users as well, I would just like this “limit” deal to be more comprehensive than “simple”.

4 Likes
20

Can you please display to the user the current compute consumed and limit? This is all we want!

7 Likes
21

I’m on the $20 Pro plan, and I’ve used Opus three times with no extra charge. I’ve also used Sonnet-4 max about 30-40% of the time. I haven’t hit a rate limit yet. My plan renewed on June 24th, so only eight days have passed. I might hit a rate limit soon, I guess.

I suspect the so-called ‘burst limit duration’ might be a monthly calculation. In the worst case, I mean that the total cost for the billing month is calculated, and if it exceeds a certain amount, you’ll get rate-limited very often. So, the ‘burst limit refill slow’ actually means it refills very slowly until the next month. This is just my guess. If I can maintain my current usage pattern for a total of 20 days without hitting rate limit, I think it’s a much better deal than the old pricing.

1 Like