Wow. guys… I think this is an absolutely terrible commercial decision. The discounted third party model pool was your entire differentiator.
Doesn’t matter how much you spin it, Composer 2.5 and Grok 4.7 are not even comparable to the Anthropic models for coding. Developers will go wherever they can use the best models at the cheapest price… and this will no longer be Cursor.
Gemma 4 and Qwen3.8 are good local models, but what size are you planning to run? The 27B variants are great for their size but are no match for Grokbot.
If I had $100k to blow, I’d probably run either Kimi or Deepseek largest variants.
And yet here’s a forum full of people that are instantly regretting unifying their accounts because they didn’t realise it was a massive downgrade.
Thanks a lot for your reply @Warren_James!
I have a VMware ESXi server with 256GB of RAM & a dozen CPU cores, but no GPU’s [yet].
I have literally no idea about any open models at all, or their sizing. All I know is what I posted that you replied to.
I use Cursor and Grok Bot heavily for coding, and I’ve just hired a couple of staff that I want to use it too.
I have the legacy Cursor subscription but I am concerned about buying the X subscription for my staff, given the hundreds of replies in this forum thread.
What would you recommend please? Thank you so much
Yeah, for NOW…
But this makes it seem pretty clear this change IS coming.
Given how Cursor handled the transition from legacy 500-request-per-month accounts to the current (pre-X) model… I’m not super optimistic about this impending change 
I haven’t done a rigorous analysis, but here’s my napkin math from the usage data I can see. My model usage hasn’t changed (mostly Grok 4.6 on Medium).
Before ($20 Cursor plan): 399M tokens per week (167M Cursor + 232M Grok Bot), using 4.33 weeks per month to convert monthly usage.
After ($30 Unified X plan):
Week 1 (actual): 383M tokens (231M Cursor + 152M Grok Bot)
Week 2 (projected): 370M tokens (359M Cursor + 11M Grok Bot)
Average: 376M tokens per week
Caveat: Token value depends on the model, so Cursor and Grok Bot tokens may not be directly interchangeable. However, my weekly totals remain similar despite very different usage splits across the two weeks, and my model usage in Cursor hasn’t changed. This suggests that adding the two together is a reasonable way to compare my overall usage.
That’s a 6% drop in total tokens despite a 50% price increase. If I’m using only Cursor + Grok Bot, I’d expect at least 600M tokens per week (399M × 1.5) to scale proportionally with the price increase. Instead, I’m at 376M, which is about 20M tokens per dollar before vs. 12.5M after, a 37.5% decline in tokens per dollar.
Yes, I now get 2x or more pure Cursor usage, but my total tokens (Cursor + Grok Bot) have gone down substantially relative to the price increase. That matches my experience of running out faster.
Thank you for the clarity, but there’s one setting am not okay with in the new unified X Subscription.
I’m from a cursor background and we have always had the option to completely turn off the credit usage to avoid unintended bills. After switching to the new plan I can see the option to set monthly credit limit, however there’s no option to turn it off completely and I just stay within included usage.
Given that third-party models are all billed at external API rate, a minor unintended switch from Grok or Composer models to forexample Fable would execute without any warnings if I have not set a credit limit for the month and I have left it to 0, I would wake up finding a whole bill of Fable cost that I never intended to use. Imgaine a Premium user at $35 finding $1000 credits used of Fable accidentally because of that accidental model switch in cursor.
Kindly bring back the option to disable credits all together, in that even if I accidentally switch to third-party model and I have not set a monthly credit limit or I have disabled credit usage, the request just fails rather than bringing me unintented bills.
I have searched my entire settings panel but the only workaround am seeing is setting a limit of $1, plus disabling the external models in cursor settings..all of which I have realized after a third-party model had already pumped several $s into expenses I never budgeted for.
Hey @Fredm
You can turn credit usage off completely. On the new plans, a monthly limit of $0 is the off switch: with a $0 limit, anything that would use credits (third-party models, or usage past your weekly included allowance) is refused instead of charged.
When no limit is set, the field shows a grey “0”, but that’s only a placeholder and an empty field means no limit. You need to type the 0 and save it.
*You’ve already set a $1 limit, so editing that same row down to 0 works too. Credits are prepaid, so it’s also worth checking that automatic top-ups are off on the same card. Turning off third-party models in Cursor Settings → Models, as you did, is a good extra safeguard!
@Colin That sounds like a very poor choice of placeholder by the Cursor team.
It gives 0 two opposite meanings, depending on whether its being shown via a placeholder or via text. One meaning ‘unlimited’ and the other meaning ‘none’.
Maybe ∞ or -1 would have been a better choice of placeholder for such a critical field.
I can easily see you getting a bunch of chargebacks for this, with people attaching a screenshot of their on-demand Cursor limit showing 0 to their bank.
Although I purchased the Cursor package, I no longer use this IDE. The official actions of this IDE are too slow. The Chinese support, agent function, and Android version have all been in the planning stage for several months but haven’t been released yet. I will not purchase XAI’s products anymore after the expiration date.
Hello, I saw that Cursor’s plans will be unified and merged with Grok’s. I have two questions:
For those, who haven’t switched to the X subscription, please DON’T! This is the worst choice!
Dear support, is there any way to switch back? I feel like I got scammed. It is my fault not to do proper research before switching, because I believed in you and your honesty. What can I do? My daily usage on GrokBot now just erased both Cursor’s and GrokBot’s weekly limit. That is something unfair.
If you’re reading this and you still have Cursor PRO, don’t change anything—it’s just a cash grab.
Sure, you’ll get a plan that’s twice as big, but two separate pools will be merged into one, so you’ll end up paying more for the exact same thing.
Local is cool, but wayyyy behind. Look at what it takes to run K3 at q8, which is honestly the minimum I’d call agentic coding. As a chatbot of course Gemma 4 is genuinely amazing, for local.
Hey system saved you dont switch unles you want pay more for the same limit pool 
Me too
I bought the x-subscription, but its still sayign I am in free, and this is now 24 hours
I used to subscribe to the $60 plan, primarily using the “Auto” model for most of the day. It worked perfectly for me, though I would occasionally run into limitations when handling certain tasks (specifically some programming-related ones).
When I switched to the $100 “Super” plan—taking advantage of a discount—I assumed I’d never need to look at another model again.
However, when performing the same tasks as before, it’s clear that the “Auto” model’s performance just isn’t the same.
Currently, I’m using “Grok 4.7” on the “High” setting to achieve the level of performance I previously had, as the accuracy I once relied on in “Auto” and “Composer” is no longer there.
I can’t explain exactly why—I lack the technical expertise for that—but I know that the changes went beyond just the pricing.
I honestly feel that the promised 2.25x increase in usage limits doesn’t actually exist.
As a result, after a week on this new subscription, I find myself considering other models once again. I’ll likely give Anthropic a try.
Yep. This is pretty much how everyone feels. All of their older model performance has significantly went down hill and you are forced to use Grok 4.7 or Other models. I find composer 2.5 basically useless now. With no Other models quota, you are forced to use Grok 4.7 because nothing else is capable. 4.7 isn’t even capable… and Other models were necessary to correct Grok and review it’s work.
That’s only the agent side. Then the pricing… Pricing has slowly been creeping up behind the scenes, every update, every new “plan”, every merge. The marketing says this is a great idea! Switch and you will have all the usage you could ever need! While in reality, they are probably giving you half of what the legacy plans used to offer, and they make it seem like they are doing you a favour.
It’s honestly sickening. Everything a moderator says, doesn’t line up with their Cursor pricing page, because they update it constantly and don’t keep it in sync.
They don’t have transparent pricing on how models are actually billed, they don’t notify their customer base about anything because they want everything to be obfuscated behind the scenes to hide it from the public.
There is definitely some shady stuff going on. We can all see it. They won’t talk about it.
Most likely they are going to lose 50% of their customer base because of this. Grok Bot is a joke and full of bugs. I would never put all my eggs into cloud computing or Grok computing. What if something breaks? What if you can’t access your own stuff? What if this takes weeks to fix? That’s not just on the company, but on you as well for having the faith that Grok would provide a dependable service.
But just like everything these days, the consumers are bug testers, the release software is full of bugs, and beta / alpha builds get pushed out to listen to the masses complain and file tickets for them. This is the way of the development world and has been in this direction for quite some time.
The worst of it is, if you actually have an issue, and you report it, you get cookie cutter AI responses for everything. No verified humans at all. For all we know, this entire product is run by a fleet of agents and they have just tailored the personalities of their bots to disguise themselves as real people.
That’s just my two cents, I could be way off base, but from my observations, and what I have read about how other devs are feeling on these forums, there is a serious product issue, and Cursor could care less… probably because they are loosing the AI race, and probably loosing money from the Legacy plans.