Cursor pricing: the X Subscription plans and my $510.70 bill

Contents

Cursor plan prices at a glance

Cursor AI pricing now runs $0, $8, $30, $100 or $200 a month, and none of those prices include Claude.

That is the short version of a big change. From 12 October 2026 Cursor stopped selling its own plans to new customers and started selling the Cursor X Subscription, officially The X Subscription, one subscription covering Cursor, Grok and X. The $20 Pro plan most people know is now a legacy plan, kept for existing customers.

C

Here is the new ladder, read off Cursor’s pricing page and its models-and-pricing docs in a US browser on 27 October 2026.

Plan	Monthly	Annual	Usage	Claude, GPT, Gemini
Free	$0	$0	Limited	Not available
Plus	$8	$80 a year	The base pool	Not available
Premium	$30	$300 a year	4× Plus	API list price, prepaid
Super	$100	$1,000 a year	15× Plus	API list price, prepaid
Ultra	$200	$2,000 a year	50× Plus	API list price, prepaid
Teams	$30, $100 or $200 per seat	—	Matches the seat’s tier	API list price + $0.25 per M tokens
Enterprise	Custom	Custom	Pooled	API list price + $0.25 per M tokens

Annual billing is ten months’ price for twelve on every paid tier, which works out to just under 17% off. The old plans gave a flat 20%.

I paid for Cursor for eleven months on the old plans, and my bill came to $510.70 on a plan advertised at $20. Most of this page is about why the new plans are a different bargain from the one I had, and who comes out ahead.

What changed, and why it happened

Two events sit behind the new price list.

SpaceX closed its $60 billion all-stock purchase of Anysphere, Cursor’s parent company, on 14 August 2026. Cursor now sits alongside Grok and X, and the FAQ on its pricing page says subscriptions are sold only through X Subscription Corp. The X Subscription is what that merger looks like at the checkout: one weekly pool of AI usage shared between Cursor and the Grok apps, with X Premium perks attached.

The second event is about models. On 28 August, OpenAI said it intends to wind down its contract supplying its models to Cursor, with a proposed shutoff date of 12 November 2026, and that it will not give Cursor its future models. Cursor’s docs still list GPT-5.6 Luna, Sol and Terra today, so GPT works in Cursor now. It works on a clock, though, and the date is a proposal rather than a certainty.

Put those together and the shape of the new plans makes sense. Cursor’s own models, Grok and Composer, are what the subscription pays for. Everyone else’s models are something you buy on top.

The plans, one at a time

Free, $0. Agent and Chat with limited usage, and no card. Cursor’s docs name it the closest match to the legacy Hobby tier, though not an identical one. Cursor does not say how big the allowance is.

Plus, $8. Agent with Cursor’s own models only: Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5. They run at fixed settings, with Fast mode off and medium effort, and Grok 4.7 capped at its 256k context window. Plus also includes Grok Chat, Voice, Imagine and Build, Grok API access and an X Premium checkmark. It does not include Auto, Cloud Agents, Automations, Bugbot, Origin, third-party models or your own API keys. Cursor’s own pricing FAQ recommends Plus for “social creators with limited AI usage”, which tells you who it is really for.

Premium, $30. Cursor labels this one “most popular” and recommends it for new AI users. It gets four times Plus’s usage and adds Cloud Agents and Automations, code hosting on Origin, code review with Bugbot, Grok Bot and access to third-party models at API list price.

Super, $100. Fifteen times Plus’s usage and priority model access at peak times. Cursor pitches it at daily Agent users.

Ultra, $200. Fifty times Plus’s usage and the earliest access to new products and models. This is the only tier that kept both its name and its price from the old list.

Teams. Each member gets a Premium, Super or Ultra seat at the individual price, and you can mix seat types on one team. Plus is not available for teams. Seats add centralised billing, a team marketplace for rules, skills and plugins, usage analytics, team-wide privacy mode and SAML/OIDC SSO.

Enterprise. Custom pricing, pooled usage across Grok and Cursor, invoicing, SCIM and the finer spend and access controls.

Feature	Plus	Premium	Super	Ultra
Agent with Grok and Composer	Yes, fixed settings	Yes	Yes	Yes
Auto (Cost mode)	No	Yes	Yes	Yes
Cloud Agents and Automations	No	Yes	Yes	Yes
Bugbot code review	No	Yes	Yes	Yes
Claude, GPT, Gemini	No	List price, prepaid	List price, prepaid	List price, prepaid
Priority access at peak times	No	No	Yes	Yes

What happened to Pro at $20?

It still exists, but you cannot buy it any more.

Cursor’s docs list Hobby, Pro, Pro+, Ultra and the Teams Standard and Premium seats as legacy Cursor plans. They “keep working as they are” for people already on them, at the same price, until those customers cancel or switch. New customers start on The X Subscription.

The docs also map each old plan to its nearest new one:

Legacy plan	Price	Closest X Subscription tier
Hobby	Free	Free
Pro	$20	Premium, $30
Pro+	$60	Super, $100
Ultra	$200	Ultra, $200
Teams Standard	$40 per user	Premium seat, $30 per user
Teams Premium	$120 per user	Super seat, $100 per user

Two lines in that document matter more than the table. First, each paid X Subscription tier “includes at least twice as much coding usage with Cursor models as the legacy plan it replaces”. Second, it “drops the included third-party model allowance”. Old Pro, Pro+, Ultra and the team seats each carried a second pool for Claude, GPT and Gemini. The new plans do not.

Switching is optional and, in Cursor’s word, permanent. There is no way back to the legacy plan afterwards.

Is Premium better value than old Pro?

It depends almost entirely on which models you use.

If you mostly let Cursor’s own models do the work, Premium is the better deal. You pay $10 more and get at least twice the usage of Grok and Composer, plus Cloud Agents, Bugbot and Origin. Weekly resets also help in a way the old monthly pools did not, because one heavy day no longer eats the whole month.

If you live in Claude, the maths turns around. Legacy Pro’s $20 bought both pools, so some Claude usage came inside the price. On Premium every Claude token is extra, at list price, from the first request. A developer who used to spend a Pro allowance’s worth of Claude now pays $30 plus that same Claude spend again as prepaid credit. For that person the new entry price is $30 plus whatever their Claude habit costs, not $30.

And if you lean on GPT models in Cursor, read the OpenAI date before you commit to anything annual.

My own eleven months would have landed on the wrong side of that line. Most of my metered usage was frontier models, not Composer.

How usage works on the new plans

The machinery is simpler than the old plans, and it sits in Cursor’s models-and-pricing docs.

One pool, weekly. Each paid plan includes one pool of usage for Cursor’s models, refreshed every week. The pool is shared across everything on the subscription: Agent in the editor, the CLI, Cloud Agents, Grok Bot and the Grok apps. A long evening chatting with Grok spends the same allowance your coding agent uses.

Cursor’s models draw from the pool. That means Grok 4.7, 4.6 and 4.5 and Composer 2.5. Different models draw at different rates.

Third-party models never touch the pool. Claude, GPT and Gemini are billed at their API list price from prepaid credit, on Premium and above only.

Extra usage is prepaid. When the weekly pool runs out you can wait for the reset, buy credit and keep going at the same per-model rates, or move up a tier. Cursor says requests are never downgraded in quality or speed.

Individual plans get Auto in Cost mode only. Cost mode bills as a Cursor-model request and draws from the weekly pool whatever model answers. The Balance and Intelligence modes, which can route to third-party models, are a Teams and Enterprise feature.

	Legacy plans	The X Subscription
Included usage	Two monthly pools: Cursor models and other models	One weekly pool for Cursor models, shared with Grok
Claude, GPT, Gemini	Included through the other-models pool	Not included; API list price from prepaid credit
Extra usage	On-demand, billed in arrears	Prepaid top-ups, optional Auto Top-up
Spend control	On-demand spend limit	Monthly additional usage limit
Billing lives at	cursor.com/dashboard/billing	grok.com/dashboard/billing

What the bill is made of now

The old plans had one trap, and the new plans remove it.

On my legacy plan, usage past the allowance was billed in arrears, a month after I spent it. That delay is how a $20 plan quietly turned into $510.70. You do not feel a meter running. You feel an invoice four weeks later.

On The X Subscription, extra usage is prepaid. You buy credit first, and both third-party models and Cursor-model overage draw it down. If the credit runs out, the extra usage stops until you add more. The surprise bill becomes a balance you can watch.

Two settings decide whether that stays true:

• Auto Top-up refills your credit when the balance drops below a threshold you set. It is convenient, and it turns prepaid back into something close to a running meter.
• The monthly additional usage limit caps how much paid extra usage you can use in a month, however much credit you hold. This is the new version of the spend limit I never set.

Both live under Additional usage credits at grok.com/dashboard/billing, not in Cursor’s own dashboard. That move alone will confuse a lot of people looking for the old Spending page.

Teams carry one extra charge. Eligible third-party model requests on Teams and Enterprise include a Cursor Token Rate of $0.25 per million tokens on top of the model’s price. Individual plans do not pay it, and Cursor’s own models never carry it. Choosing regional data residency adds a 10% uplift on eligible model pricing on any plan.

My Cursor invoices, line by line

All of this is history now, but it is real history, and it is the reason this page exists. I paid for Cursor on the old plans for eleven months and then cancelled. Here is the whole of it, from the Stripe invoices rather than from memory.

	Amount
Subscriptions across eleven months	$340.00
Metered usage on top	$170.70
Total	$510.70
Worst single month	$159.22
Usage as a share of the total	33%

I joined at $20 a month. In October 2025 I upgraded to Pro+ at $60, specifically to stop the overages. It did not work. A bigger allowance moved where the line was without changing what happened when I crossed it.

The month that made the point was October 2025: $159.22, on a plan advertised at $20, with roughly half of it unplanned. Nothing was broken. I had a heavy month and the bill arrived a month later.

Run the same eleven months through the new plans and two things change. The arrears surprise goes away, because extra usage would have been prepaid. But every dollar of frontier-model usage would have been extra from the first request, with no included allowance to absorb any of it. For someone who used Cursor the way I did, the new list is easier to predict and probably no cheaper.

Did Cursor cause the bill shock?

Mostly not, and that still matters, because “Cursor is expensive” was the wrong lesson then and it is the wrong lesson now.

Cursor charged the model’s API rate with no markup on individual plans. It published the rates. It offered a spend limit on every individual paid tier. That spend limit would have capped my entire problem at whatever number I chose, and I never set it.

What was fair to put on Cursor was the shape: a headline price that was a floor rather than a price, and arrears billing that put the feedback a month behind the behaviour. The X Subscription fixes the second half of that. The first half is still there, just moved. The headline number is now $30, and for anyone who wants Claude it is still a floor.

Per-model token rates inside Cursor

Since third-party models are billed at list price, the model you pick is the price you pay. A sample from Cursor’s published rate card, per million tokens:

Model	Input	Cache read	Output
Composer 2.5	$0.50	$0.20	$2.50
Grok 4.7	$2.00	$0.50	$6.00
Claude Sonnet 5.5	$2.00	$0.10	$10.00
Claude Opus 5.5	$4.00	$0.20	$20.00
GPT-5.6 Sol	$4.00	$0.40	$20.00
Gemini 3.8 Flash	$0.75	$0.075	$3.75

The Composer and Grok rates apply only once you have used up the weekly pool. Until then those models cost nothing extra.

Two things jump out. Composer 2.5 is a quarter of Claude Sonnet 5.5’s output price, which is exactly the bet Cursor is making with these plans. And the long-context and Fast variants are price multipliers: Grok 4.7 with its 500k window costs $4 in and $12 out, twice the standard rate. Turning on the big context window is a pricing decision.

Max Mode now exists only on legacy request-based plans, where it adds 20% to the model’s API rate in exchange for a larger context window.

Five ways to spend less on Cursor

None of these are clever. Most of them are things I did not do.

Tactic	Where it lives	What it changes
Set the monthly usage limit	grok.com billing	Caps paid extra usage outright
Leave Auto Top-up off	grok.com billing	Credit stops instead of refilling
Default to Composer or Grok	Model dropdown	Draws from the pool, not your credit
Treat long context as a purchase	Per request	Avoids the 2× long-context rate
Check the pool mid-week	Editor settings	The reset date is shown there

Set the monthly additional usage limit. It caps paid extra usage however much credit you hold. On a legacy plan, set the on-demand spend limit instead. If you do nothing else on this list, do this one.

Leave Auto Top-up off until you know your rhythm. Prepaid credit that stops when it runs out is the best protection the new plans give you. Auto Top-up quietly removes it.

Default to Cursor’s own models. On The X Subscription this is the biggest lever on the bill, because Composer and Grok come out of the pool while every Claude request comes out of your credit. Reach for Claude when the task needs it, not out of habit.

Treat extended context as a purchase. A 500k window at twice the rate is a fair trade when you need it, and waste when the answer was three files away.

Check usage mid-week. Your remaining usage and next reset date sit in the editor settings and under Usage at grok.com/dashboard. A weekly pool rewards pacing.

One feature that moved tiers: Bugbot. Cursor’s automated code review is now part of Premium and above, and Plus does not get it at all. On the old individual plans it drew on usage-based billing. If code review is the reason you want Cursor, $8 is not your entry price.

Cursor Teams, priced per seat

Team pricing got simpler and, at the low end, cheaper.

Each member gets a Premium seat at $30, a Super seat at $100 or an Ultra seat at $200, with the same weekly usage as the matching individual plan. You can mix seat types on one team, so two heavy agent users can sit on Super while everyone else stays on Premium.

For ten developers on Premium seats, that is $3,600 a year at list, against $4,800 for ten legacy Standard seats at $40. Before usage, the new team plan is cheaper.

Usage is where the arithmetic changes. Team seats include no third-party model allowance, eligible third-party requests carry the $0.25 per million token Cursor Token Rate, and regional data residency adds 10%. On Teams, admins manage one shared prepaid balance and one spending cap for the whole team. Set the cap before you hand out seats.

On Teams and Enterprise, Cursor Router can also pick between Cursor’s models and third-party models for each request in Balance or Intelligence mode. A route to a third-party model bills as third-party usage. That is useful, and it is also a way for Claude spend to arrive without anyone choosing Claude.

How Cursor pricing compares

Against the tools I have tested on identical work, Cursor now has the cheapest paid entry and the most conditions attached to it.

Tool	Entry price	Free tier	Billing shape
Cursor	$8/mo (Cursor models only)	Yes, evaluation-grade	Weekly pool, plus prepaid third-party usage
Claude Code	$20/mo	No	Subscription, or metered API
Google Antigravity	Free	Yes, does real work	Free, or a Google One plan
Codex	Bundled with ChatGPT	Yes	Included in the subscription
Lovable	$25/mo	5 credits a day	Credits, which expire

The comparison people used to make was “$20 versus $20”. Now it is “$8 for Grok and Composer, or $30 plus Claude at list price, versus $20 for Claude Code with Claude included”. If Claude is the model you want, Claude Code’s $20 buys more Claude than any Cursor plan, because no Cursor plan includes any.

And Antigravity is still the only one of these whose free tier will carry real work indefinitely.

Matching a Cursor plan to how you code

If you mostly…	Pick	Budget for
Want to try the editor	Free	Nothing
Use Grok and Composer, lightly	Plus	$8 a month
Want Claude available at all	Premium	$30 plus your Claude spend
Run agents every day	Super	$100 plus third-party spend
Run several agents all day	Ultra	$200 plus third-party spend
Are on legacy Pro and live in Claude	Stay put	Your current $20

Free — if you are evaluating. It asks for no card and it will tell you within a week whether the editor suits you.

Plus, $8 — only if Grok and Composer are enough for you. It is a real coding agent at a very low price, but it has no Claude, GPT, Gemini, Bugbot, Cloud Agents or Auto. Treat it as a cheap way to use Cursor’s own models, not as a cheap Cursor.

Premium, $30 — the sensible default for one developer, and the first tier where Claude is available at all. Budget $30 plus your expected third-party spend, and set the monthly usage limit on day one.

Super, $100 — for daily agent users who have watched themselves run out of Premium’s weekly pool. Buy it for the usage you have measured, not to escape surprise bills. Buying a bigger allowance to stop overages is the mistake I made.

Ultra, $200 — for people running several agents or Automations all day. At this price, compare it against your own month of metered usage before you commit.

Already on legacy Pro, Pro+ or Ultra? If you lean on Claude or GPT, staying put may be the better deal, because your plan still includes a third-party pool. If you mostly use Composer and Grok, Premium or Super gives you at least twice the usage. The switch is permanent, so check a month of your own usage first.

Teams — Premium seats for most people, Super seats for the heavy users, and a team spending cap before anything else.

Inside Cursor’s free Hobby plan

“Is Cursor free” is one of the most common things people search alongside the price, and the answer deserves more than yes.

The free plan asks for no card. Cursor’s pricing page lists limited Agent requests and limited Chat requests against it, and it does not publish what “limited” means in requests or dollars. You size it by hitting the end of it, like several rivals.

I do have one measured data point, from the old Hobby tier in August. The subscription-tracker build described below met six requirements in 581 lines in under a minute and cost nothing. So the free tier is not decorative.

What it will not do is the thing that produced my $510.70: all-day iterative work against frontier models. Free includes no third-party models, so on Free you are working with Cursor’s own models only.

Cursor’s pricing page does not advertise a free trial of the paid tiers, so treat the free tier as the trial.

Cursor for students: no discount, yet

The web is fairly confident about a thing that does not appear to be true.

Several guides list a Cursor student discount as an established offer, sometimes with terms attached. Cursor’s own students page, checked again on 27 October, does not describe one. It says anyone can get started for free, and tells students to watch for promotions at on-campus and online events, with a campus newsletter for early notice.

That is not a discount programme. It is the free tier everyone has, plus the chance of a promotion with no published terms, no stated eligibility and no verification flow. If you are a student, Plus at $8 is now the cheapest paid way in, with the model limits described above.

How Cursor’s pricing got here

A good share of the people searching for Cursor’s price are really asking what happened to it, and the history explains the plans.

	Before mid-2025	Mid-2025 to Oct 2026	Since Oct 2026
You bought	A count of requests	An amount of model spend	A weekly pool of Cursor-model usage
Third-party models	Counted as requests	Included in a second pool	Extra, at list price
Overspend showed up	When the quota ran out	On next month’s invoice	As prepaid credit running down
Entry paid price	$20	$20	$8, or $30 for third-party models

Cursor originally sold requests, a monthly quota where 500 fast premium requests meant 500 of them. In mid-2025 it moved to dollar-denominated usage, so your plan bought an amount of model spend rather than a count of actions. That change was contentious, because a request is a unit you can hold in your head and a dollar of frontier-model usage is not.

The X Subscription is the third version in about eighteen months. It keeps the dollar pricing for third-party models, swaps the monthly allowance for a weekly one, and makes the house models the thing the subscription pays for. Whatever you think of it, a price list now on its third version since early 2025 is a reason to think twice before paying for a year up front.

One real build, priced on Cursor

Abstract rates are hard to weigh, so here is one concrete data point from the same brief I have given seven AI coding tools.

In August I asked Cursor to build a single-page subscription tracker with six specific requirements: add a tool with a name, cost and billing cycle; show annual plans as a monthly equivalent; a running total pinned to the top; a cancelled state that leaves the total but stays visible; localStorage persistence; and a dark theme.

It met all six in under a minute, writing 581 lines across three files, on the free tier, for nothing. It was the fastest of the seven tools I ran that brief through.

That is the case for Cursor’s pricing that my invoice story hides. A single well-specified task, done in one pass, costs about nothing. My $510.70 came from eleven months of open-ended, all-day work, which is a different activity in the same editor. If your use looks like discrete jobs, Cursor can be cheap. If it looks like living in the thing, size the plan to your usage and set the cap.

Final word

Cursor’s new pricing is cheaper to start, easier to predict and narrower than what it replaced.

The $8 entry price is real, and prepaid extra usage removes the month-late invoice that cost me most. Premium gives at least twice the old Pro’s usage of Cursor’s own models for $10 more. If Composer and Grok do your work, that is a better deal than the one I had.

But the new plans are built around Cursor’s own models, and everything else is extra. No Cursor plan you can buy today includes Claude, and OpenAI has proposed pulling its models from Cursor on 12 November. If you came to Cursor for frontier third-party models, your real price is $30 plus your usage, and you should treat it that way from day one.

If you take one thing from this page, take the boring one. Pick the tier for the models you actually use, and set the monthly usage limit before you write any code.

If the bill is the reason you are reconsidering, here is what each rival cost for the same build. The full Cursor review is where these invoices came from, Cursor alternatives covers what to move to, and Claude Code vs Cursor and Cursor vs Codex are the two switches people actually make.