Disclosure: some links on this site may be affiliate links. If you buy through one, we may earn a commission at no extra cost to you. It never changes what we write or how we rank things. Details.

Cursor pricing explained: the weekly usage pool, prepaid credit and per-model rates (checked November 2026)

Updated November 14, 2026 · Facts come only from the vendors' official pages listed under Sources. We have not run paid accounts on these products for this page.

Cursor is now sold as part of The X Subscription, which its docs describe as "one subscription for Grok, Cursor, and X". Every paid plan includes a weekly pool of usage for Cursor's own models, shared with the Grok apps. Third-party models such as Claude, GPT and Gemini are not in the pool at all: on Premium and above they are paid from prepaid credit at API list price from the first request. The pricing page gives pool sizes only as multiples of the cheapest plan. Below is what Cursor's pricing page and docs said on November 14, 2026, and a worked example of what one agent session costs on each kind of model.

The short answer

• Plans: Free ($0), Plus ($8/month), Premium ($30), Super ($100) and Ultra ($200). Teams seats cost $30, $100 or $200 per user per month. Annual billing is $80, $300, $1,000 and $2,000 a year for the paid individual plans. Prices exclude tax.
• The weekly pool covers Cursor models only: Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5. It is shared with Grok Chat, Imagine, Voice and Grok Bot, it refreshes weekly, and unused usage does not roll over.
• Pool sizes are relative. Premium has 4x Plus's pool, Super 15x and Ultra 50x. Cursor publishes no dollar or token figure for Plus itself.
• Third-party models are never in the pool. On Premium and above they cost API list price from the first request, paid from prepaid credit. Free and Plus can't use them.
• Past the pool, Cursor models keep working from prepaid credit at their per-token rates. Credit is bought up front, by hand or with Auto Top-up (off until you turn it on), and expires 365 days after purchase.
• Teams pay a Cursor Token Rate of $0.25 per million tokens on most third-party models. Individual plans don't.

What each plan includes

Notes on the table:

• Plus has fixed settings. Grok 4.7 runs with the 256k context window, and Plus has no Auto, Cloud Agents, Automations, Bugbot, Origin, third-party models or your own API keys.
• Cursor's own advice: Premium for new Agent users, Super for daily Agent users, Ultra for power users running several agents or Automations. The pricing page says Plus "is recommended for social creators with limited AI usage."
• X features are part of the same plans (the pricing page lists things like ad-free scrolling on Premium). This page covers only the Cursor side.
• The Grok API is included on paid plans but billed separately at standard model rates; it doesn't draw from the pool.
• Changing plans: upgrades apply right away and downgrades at the end of the billing period. The pricing page says to cancel at least 24 hours before renewal to avoid another charge.

The weekly pool

Cursor's docs put it this way: "The pool is shared across everything on your subscription: Agent in the editor, CLI, Cloud Agents, Grok Bot, and the Grok apps. It refreshes weekly, so one heavy day doesn't use up your month."

• What draws from it: Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5, and Auto in Cost mode, which "bills as a Cursor-model request and draws from your weekly pool whatever model serves it." Cost mode is the only Auto mode on individual plans.
• What doesn't: third-party models, the Grok API, and, on individual plans, requests made with your own API key (your provider bills you directly).
• Where to see it: Usage at grok.com/dashboard shows how much of the pool you've used, the next reset date and your prepaid credit. Your editor settings show the same remaining usage.

How big is it?

Cursor doesn't say. The plans are described only relative to Plus, and the Team pricing docs give the same ratios from Premium's side: a Super seat has 3.75x Premium's pool and an Ultra seat 12.5x.

That is still enough to compare what each plan charges per unit of pool, even without knowing the unit:

Per unit, Ultra's pool costs half what Plus's does. This counts only the included pool; Premium and above also add features and third-party model access.

How fast each model uses it

The docs say "different models draw from the pool at different rates", and the top-ups page says Cursor models past the pool are billed "at the same per-model rates as included usage." If the pool is drawn at those rates, Composer 2.5 lasts several times longer than Grok 4.7: in our example below, the same session costs about a quarter as much on Composer 2.5 with nothing cached, and a little over a third as much with 90% cached.

Prepaid credit (top-ups)

Top-ups add prepaid credit to your subscription. From Cursor's help pages and the X Subscription docs:

• What it pays for: third-party models from the first request (Premium and above), and Cursor models after your weekly pool runs out.
• When you're charged: when you buy credit or when an automatic top-up runs, "not in arrears."
• Auto Top-up is off until you turn it on. You choose the balance that triggers it and the amount it adds.
• Expiry: credit doesn't reset with the weekly pool. Each top-up lasts until used or until it expires, "currently 365 days after purchase."
• Spending cap: a "monthly additional usage limit" caps how much paid extra usage you can consume each month, "regardless of how much credit you hold." Adding credit doesn't count toward it. Cursor notes that enforcement "is not instant, so usage can briefly exceed your spend limit."
• With Auto Top-up off, extra usage stops once your weekly pool and your prepaid credit are both used up.
• Credit grants from promotions are separate from prepaid credit, are applied automatically, and show their own expiry date.
• Where: Additional usage credits at grok.com/dashboard/billing. The pages we read give no minimum top-up amount.

When the pool runs out, Cursor says "requests are never downgraded in quality or speed": you wait for the reset, use credit, or upgrade.

Per-token rates

Cursor models (billed from credit past the pool)

The docs say these rates "apply when you buy additional usage beyond the weekly pool." No cache-write price is listed for Cursor models.

All prices are US dollars per million tokens. For Grok 4.7, input above 256k tokens is billed at 2x the standard rates (up to 500k), Fast mode at 2x, and Fast mode with long context at 3x.

Third-party models (Premium and above, from credit)

These are the models shown by default. The docs table also lists 43 older or hidden-by-default models.

All prices are US dollars per million tokens. Dated changes and surcharges the table notes:

• Gemini 4 Carbon: introductory pricing through November 30, 2026. From December 1: $4 input, $0.20 cache read and $20 output per million.
• GPT-5.6 Sol: "Promotional pricing through November 21, 2026." The table doesn't say what it costs after that.
• Claude Haiku 5.5: requests above 100k input tokens are billed at 5x the standard rates.
• GPT-5.6 Luna, Terra and Sol: long context up to 1M tokens at 2x input pricing; Fast mode at 2x.
• Claude Sonnet 5.5 and Opus 5.5: up to 1M tokens at the same per-token rates, with no long-context surcharge.
• Regional data residency: opting in adds 10% on eligible models.
• Claude Fable 5.2 requires data retention approval for Enterprise, Teams, and individual customers with Privacy Mode on.

Worked example: one agent session

We use the same made-up session as in our Copilot and Devin Desktop articles: 30 model calls, each sending 40,000 input tokens and getting 1,500 back, so 1.2 million input tokens and 45,000 output tokens. We show it with nothing cached and with 90% of each call's input read from the cache. Cache writes are left out, so treat the cached column as a floor.

Cursor models, priced at the rates above. Inside your weekly pool you don't pay these amounts; they are what the session costs from credit once the pool is used up.

Third-party models, paid from credit from the first request:

How to read it:

• Caching saves less on Cursor's own models. Grok 4.7's cache-read price is a quarter of its input price, so 90% caching cuts this session by about 61%, and Composer 2.5's (40% of input) by about 45%. On Claude Sonnet 5.5, where cache reads cost 5% of input, the cut is about 72%.
• Fast variants cost a lot more. Grok 4.7 Fast is exactly double Grok 4.7. Composer 2.5 Fast is six times Composer 2.5 with nothing cached.
• Third-party models cost the same here as elsewhere. The uncached session on Claude Sonnet 5.5 is $2.85 in Cursor, the same as 285 credits on GitHub Copilot and the same as Devin Desktop's extra usage. What differs is the packaging: Copilot Pro's $10 a month includes 1,500 credits ($15 at list price) that work on these models, while Cursor's $30 Premium includes none, so every third-party request is paid from credit on top of the plan.
• On Teams, add the Cursor Token Rate for eligible models: $0.25 per million on input, output and cached tokens. This session is about 1.25 million tokens, so about $0.31 more. That is 11% on Claude Sonnet 5.5 with nothing cached ($3.16), but 39% with 90% cached ($1.11), because the rate is flat while cached tokens are cheap.

To try your own token counts and plans on Cursor, GitHub Copilot and Devin Desktop side by side, use our AI coding cost calculator. It uses the rates above, adds the Cursor Token Rate on Teams seats, and shows Cursor-model usage as a range, since the pool's size isn't published.

Teams

• Seats: Premium ($30), Super ($100) or Ultra ($200) per user per month, each with the same weekly pool as the matching individual plan. One team can mix tiers. Plus isn't available for teams.
• Free seats cost $0 and have no team usage allowance; that member's usage draws from their own personal plan, or a small free allowance if they have none.
• Pools are per seat. Seats don't share pools, and usage doesn't move between members.
• One shared prepaid balance pays for all third-party usage and for Cursor-model usage past a seat's pool. Only admins buy credit, and admins set one spending cap for the team, either a fixed amount or an amount per seat.
• Billing is per active paid seat. Upgrading a seat applies right away with a pro-rated charge; downgrades apply at the end of the billing period.
• The Cursor Token Rate ($0.25 per million tokens) applies to eligible third-party requests, including when Cursor Router (Balance or Intelligence) routes to a third-party model and when you use your own API key. It doesn't apply to Cursor models, to Cost mode requests, or to third-party models whose cache-read price is "strictly below $0.10 per million tokens across every pricing tier." By our reading of the table, that exempts Claude Haiku 5.5, GPT-5.6 Luna, Gemini 3.8 Flash and Gemini 4 Hydrogen among the default models; Claude Sonnet 5.5 and Gemini 4 Argon list exactly $0.10, so they pay it.
• Cursor Router (Teams and Enterprise only) has Balance and Intelligence modes that bill at the routed model's list price, so a route to a third-party model is third-party usage from credit.
• Enterprise is custom-priced and adds pooled usage across the team, invoicing and SCIM.

Five developers on Premium seats would be $150 a month before any credit.

Legacy plans (Pro, Pro+ and older Teams seats)

Customers who subscribed before The X Subscription keep their plans unless they change or cancel. Switching is optional, but it is permanent and can't be undone.

A Start plan (India only, ₹649/month) maps to Plus.

• How legacy plans work: two monthly pools, Cursor Models and Other Models, with on-demand usage billed in arrears after they run out. Pro, Pro+, Ultra and Teams include the Other Models pool, so third-party models are part of the included usage.
• What switching changes: Cursor says each new paid plan "includes at least twice as much coding usage with Cursor models as the legacy plan it replaces", but it "drops the included third-party model allowance." Your old subscription is cancelled with a prorated refund, any on-demand usage not yet invoiced is billed at the switch, and billing moves to grok.com.
• Reading those rules: if most of your work runs on Grok or Composer, the new plans give at least twice the Cursor-model usage. If most of it runs on Claude, GPT or Gemini, switching gives up an included allowance and every third-party request is paid from credit.
• Max Mode is available only on legacy request-based plans, billed at the model's API rate plus 20%.

Ways to keep the bill down

From Cursor's help pages:

• Use Cursor models for routine work. They draw from the pool, and Composer 2.5 has the lowest per-token rates of the four.
• Leave Auto Top-up off, or set a monthly additional usage limit, so extra usage stops at a figure you chose.
• Skip Fast variants unless you need the speed; they cost 2x or more.
• Watch the long-context thresholds: Grok 4.7 above 256k input tokens bills at 2x, and Claude Haiku 5.5 above 100k at 5x.
• Check what subagents run. Cursor notes that subagents "can also run a named third-party model while the picker still shows Auto, Grok, or Composer"; those requests bill as third-party usage.
• Use your own API key on an individual paid plan for third-party models if you already have one: those requests don't draw from the pool or credit, and your provider bills you directly. Plus doesn't allow your own keys.

Which plan fits

These follow from the published rules above, not from using the product.

• Light use on Grok and Composer only: Plus at $8 is the cheapest paid pool, at fixed settings.
• Starting with agents and wanting Claude or GPT now and then: Premium, plus a small credit balance with a monthly cap.
• Daily agent work mostly on Cursor models: Super or Ultra. The price per unit of pool falls from $7.50 on Premium to $6.67 on Super and $4.00 on Ultra.
• Already paying for X Premium or Grok: Cursor comes in the same subscription.
• A team with a few heavy users: Premium seats for most people, Super or Ultra for the heavy users, free seats for occasional ones, and one shared credit balance for third-party models.

Might not fit if

• Most of your work is on Claude, GPT or Gemini. The plan price buys no third-party usage on The X Subscription. GitHub Copilot's monthly credits do cover them; see GitHub Copilot AI credits explained.
• You need your allowance as a number. Cursor publishes pool sizes only as multiples of Plus. Devin Desktop doesn't publish its quota sizes either; see Windsurf (Devin Desktop) quotas explained.
• You want Cloud Agents, Bugbot or third-party models on the cheapest plan. They start at Premium.

To see Cursor's plans next to GitHub Copilot and Devin Desktop, read our Cursor vs GitHub Copilot vs Windsurf comparison.

Check the current numbers

• Cursor pricing
• Models and per-token rates
• Usage and limits
• Usage-based charges and top-ups
• Team pricing
• Legacy Cursor plans

Sources

1. Cursor pricing page — https://cursor.com/pricing, checked November 14, 2026
2. Cursor docs: Models & Pricing (per-model rates) — https://cursor.com/docs/models-and-pricing, checked November 14, 2026
3. Cursor help: Usage and limits — https://cursor.com/help/models-and-usage/usage-limits, checked November 14, 2026
4. Cursor help: Usage-based charges (top-ups) — https://cursor.com/help/account-and-billing/overages, checked November 14, 2026
5. Cursor docs: Team pricing — https://cursor.com/docs/account/teams/pricing, checked November 14, 2026
6. Cursor docs: Legacy Cursor plans — https://cursor.com/docs/account/pricing/legacy-cursor-plans, checked November 14, 2026
7. X Subscription docs: Prepaid balance and automatic top-ups — https://docs.x.ai/x-subscription/prepaid-balance, checked November 14, 2026