Cursor Pricing (2026): $20 Buys Two Usage Pools

Cursor pricing now starts at $0 for Free and $8 per month for Plus with Grok models only. If you mainly use Anthropic Claude, OpenAI GPT, or Google Gemini in Cursor, the new entry point is Premium at $30 per month plus separately purchased model usage. Paying for a higher tier increases your included Grok allowance; it does not buy an included Claude or GPT balance.

Since October 12, 2026, Cursor has been available to new customers through The X Subscription, which bundles Cursor, Grok, and X benefits. Existing subscribers can keep their current plan or switch as the option reaches their billing settings. The xAI announcement describes a rollout over the following days. Prices and billing rules below were verified on October 14, 2026 against Cursor's pricing page, its model-pricing docs, and xAI's pricing page.

Cursor pricing at a glance

Choose by model preference first. Plus is the low-cost paid option when Grok handles your work. Premium is the minimum new subscription for third-party models, cloud agents, and automations. Super and Ultra make sense when you need more included Grok usage.

These are the monthly prices on the current pricing page, before applicable taxes and fees. The public menu does not print annual equivalents for the new tiers, so the old annual discount should not be carried forward.

Teams is a governance purchase as well as a usage plan. Enterprise is the route for pooled organizational usage, invoicing, and advanced controls. Neither makes separately billed third-party work disappear.

If you already pay for it: Pro, Pro+ and the previous Ultra

Pro and Pro+ have left the new-customer menu; existing subscriptions have not been forcibly replaced. xAI says existing Cursor subscribers can retain their plan, price, and features. You can instead switch to The X Subscription and receive a prorated refund for the old plan when the switching option appears in your account. Existing-subscriber terms.

The model-pricing document still prints this ladder and describes separate monthly pools for Cursor Models and Other Models. It does not print a fixed dollar value for either pool. Its India-only Start plan also remains in that older document but is absent from the new subscription menu; do not assume it is a new-customer option from that page alone.

Ultra keeps its name and advertised monthly price, but the new Ultra is a different offer. The old document describes included third-party usage; the new subscription explicitly excludes it. A familiar plan name is not a promise that your billing terms survive a voluntary switch.

There is another unresolved documentation mismatch for existing business accounts. The older model docs still describe a $0.25 per million tokens Cursor Token Rate on third-party work for Teams and Enterprise, including bring-your-own-key usage. The new subscription pages advertise third-party usage at API list prices. Do not add the old surcharge to a new-subscription estimate as though the pages agree; an existing organization should confirm its retained or migrated terms in billing before switching.

For a developer whose current plan includes third-party usage, keeping it may be preferable to moving into a Grok-focused allowance. Compare your actual included usage and extra charges with the new plan's subscription plus top-ups before accepting the migration offer.

Why the old break-even points no longer work

The new usage multipliers measure the shared Grok allowance against Plus. The plan cards say “more usage,” but the comparison table labels the allowances 4x Plus, 15x Plus, and 50x Plus. These are the advertised total ratios to the same baseline, not successive multipliers from the next cheaper tier. Current allowance comparison.

The allowance covers Grok models across Cursor and Grok. Using Grok for chat, voice, images, or Bot work draws on the shared capacity too. It is not a dedicated coding allowance that remains untouched while you use the other bundled products.

The pool refreshes weekly, according to both current pricing pages. If you reach the limit, you can buy top-ups or wait for the refresh. Product costs vary, and neither page publishes the baseline as a fixed token count, request count, or dollar credit. That prevents a reliable conversion from the multiplier to a number of coding tasks.

For a Grok-heavy developer, a larger tier can address a recurring capacity shortage. For a Claude-heavy developer, the same upgrade leaves Claude usage separately billable. The relevant question is whether you need more Grok capacity, not whether your total model bill looks large.

What one agent task can cost

An agent task has a variable model charge, not a flat request price. Tokens are the units of text a model processes. The model rate table prices input, output, and cached context separately, all per million tokens:

• Claude Sonnet 5.5: input $2, cache writes $2.5, cache reads $0.1, output $10.
• GPT-5.6 Terra: input $2, cache writes $2.5, cache reads $0.2, output $12.
• Gemini 4 Argon: input $2, cache reads $0.1, output $10; no separate cache-write price is printed in its row.

Your top-up spending depends on the actual tokens charged in each category. A task that repeatedly reads context or generates lengthy edits can cost more than a short question. A plan multiplier does not discount those listed third-party rates.

For the same Claude or GPT workload, the new monthly budget is $30 plus that workload's charges on Premium, $100 plus those charges on Super, or $200 plus those charges on Ultra, before tax and any additional Grok top-ups. The subscription component changes; the third-party work remains an extra purchase.

A developer using only Grok can instead stay at $8 on Plus while within the shared allowance, provided Plus has the features they need. Grok users who need cloud agents and automations start at $30 on Premium. Their bill can remain the subscription price while included capacity lasts. Plan features and usage terms.

Cursor's free plan: who never needs to pay

Free is the current $0 entry point. It includes limited Agent and Chat requests, with no fixed numerical allowance printed on the current plan card. It suits evaluation, occasional coding, or work that can pause when the allowance runs out. Free plan.

You may never need to pay if Cursor is a secondary editor and you use it for narrow, interruptible tasks. When a limit repeatedly blocks working time, compare Plus for Grok-only use with Premium for third-party access and cloud features.

Do not carry forward the former Hobby plan's Tab promises or the old India-only Start recommendation to the new menu. The new plan cards describe a different subscription, and the entry-tier decision now depends heavily on whether Grok models cover your work.

Hidden Cursor costs and billing rules

The main billing risk is treating model access as included model usage. Premium and above let you select third-party models in Cursor, but those plans do not include a third-party allowance. Billing distinction.

Top-ups can turn a plan into a variable bill

Claude, GPT, and Gemini usage is purchased separately through top-ups at API list prices. Grok top-ups are a separate reason for spending beyond the subscription when the shared allowance runs out. The current pricing FAQ says spending caps are available; set a budget before making expensive agent work routine.

A developer comfortable with Grok can stop at the included limit and wait for the weekly refresh. A developer who chooses Claude from the outset should expect third-party usage charges from the outset, rather than waiting for an included Claude balance to run out.

Teams: per-seat prices, SSO and admin controls

Teams seats use the same prices as individual Premium, Super, and Ultra: $30, $100, and $200 per user per month. xAI's announcement explicitly gives that equivalence and lists SSO, admin controls, consolidated billing, usage analytics, and SOC 2 compliance. SSO means staff can sign in through the organization's identity provider. xAI's Teams terms.

Cursor's feature comparison specifies SAML/OIDC SSO, centralized administration, team-wide privacy settings, custom Chat retention, an internal rules and plugins marketplace, and the ability to mix seat types. Give a developer a larger seat for the included Grok capacity they need, rather than buying the largest tier for everyone.

Enterprise has custom pricing and adds pooled usage, invoice billing, SCIM seat management, granular and group spending controls, access controls, audit logs, and service accounts. SCIM automates account provisioning through your identity system. Choose Enterprise when those controls are requirements, not merely because the organization has more developers.

Included model usage is not an external API wallet

Grok API access is included in paid plans; Grok API usage is not included in the shared pool. The pricing FAQ says API calls are billed separately at standard model rates. A developer using Grok inside Cursor and an application calling the Grok API therefore has separate usage budgets. API billing rule.

The same budgeting discipline applies to third-party work: the new subscription's Grok capacity does not offset the Claude, GPT, or Gemini top-ups you buy in Cursor.

Switching, renewal and taxes

Existing subscribers can keep their current terms or switch for a prorated refund of the old plan. The migration offer is not a general promise that all consumed usage is refundable. Check the offer shown in account billing settings before switching. Migration terms.

The new pricing page says to cancel at least 24 hours before renewal to avoid further charges. Advertised subscription prices exclude applicable taxes and fees. It does not publish the previous annual equivalents, so do not assume an old prepaid price applies to a new tier. Current price conditions.

Cursor versus GitHub Copilot, Claude Code, and Devin

Compare the included model allowance and your expected extra spending, not just the subscription label. A Grok-focused bundle and a subscription that includes third-party coding usage can produce different bills for the same developer.

GitHub Copilot: compare included coding credits

GitHub Copilot supports multiple editors and combines code completion, next-edit suggestions, model selection, and agent features. Its plans page publishes included AI credits and options to buy more.

Copilot deserves a comparison when keeping your existing editor and an included coding-credit budget matter more than the Grok bundle. Cursor is the stronger fit when its coding environment and shared Grok usage are the products you want. The deeper workflow decision is covered in Cursor vs GitHub Copilot.

Claude Code: one subscription shared with Claude

Claude Code is included in paid Claude plans and shares usage limits with the wider Claude product. That is a different proposition from buying Claude usage separately inside a new Cursor subscription. Claude's plan terms.

For someone primarily choosing Claude for terminal work, compare the native subscription's included allowance with Cursor Premium plus Claude top-ups. Cursor still earns consideration if its editor and multi-model workflow are worth the separate bill. See Claude Code vs Cursor for the workflow decision beyond price.

Devin Desktop: the current Windsurf pricing destination

Devin Desktop is the product offered by the pricing destination reached from Windsurf. Its paid plans include agent allowances, model choice, cloud access, and extra usage at API pricing. Its public overview does not assign the individual allowance a fixed numerical task count. Current Devin pricing.

Choose Devin when its desktop and cloud workflow suits the job. Compare how the included allowance treats your preferred model before assuming its base plan is the whole cost. The product comparison is in Windsurf vs Cursor after Devin Desktop.

Which Cursor plan should you choose?

Choose the lowest tier that supports your models and workflow, then let recurring usage justify an upgrade. A larger Grok allowance is not valuable when separately billed third-party work causes the expense.

Choose Free for evaluation, occasional coding, or a secondary editor when waiting at a limit is acceptable.

Choose Plus for light paid use when Grok models are sufficient and you do not need the features reserved for Premium. Keep the shared allowance in mind if you also use Grok outside Cursor.

Choose Premium when you need Claude, GPT, or Gemini access, cloud agents, or automations. Budget the subscription and third-party top-ups as separate costs.

Choose Super or Ultra when a sustained Grok workload needs the larger shared allowance. Super also advertises priority access at busy times; Ultra adds earliest access to products and models. Neither includes a third-party model credit balance.

Choose Teams when centralized billing, SSO, privacy settings, and usage administration are required. Mix seat tiers to fit actual demand. Choose Enterprise for pooled usage and the advanced controls listed above.

1. 1

   Start with your model mix

   Use Free to evaluate the workflow. If Grok does the job, consider Plus; if you require third-party models or cloud features, start the new-plan comparison at Premium.

2. 2

   Separate allowance from top-ups

   Track shared Grok consumption and third-party charges independently. Set spending caps and account for Grok activity outside Cursor before deciding that the coding allowance is too small.

3. 3

   Compare before switching

   If you already subscribe, compare the retained plan's included model usage with the new plan plus top-ups. Review the prorated-refund offer in billing settings before moving. Upgrade a new subscription only when its additional capacity or features solve a recurring need.

Cursor pricing FAQ

How much does Cursor Pro cost?

Pro is an existing-subscriber plan rather than the current new-customer entry point. Its retained price and the previous plan ladder are in “If you already pay for it” above. New customers choose from Free, Plus, Premium, Super, and Ultra.

Is the Cursor $20 plan unlimited?

The previous Pro offer did not provide unlimited model usage. For new subscriptions, included Grok usage is limited and shared with Grok, while third-party usage on Premium and above is purchased separately. The older plan terms are covered in the existing-subscriber section.

Is Cursor free?

Yes. Free costs $0 and provides limited Agent and Chat requests. The current plan card does not promise a fixed numerical allowance, so it is best for work that can pause at the limit.

Does Cursor offer a student discount?

Cursor's current Students page points to free access and promotions at campus and online events. It does not promise a standing student subscription discount. Check the live Students page for an offer you qualify for.

What is Cursor's refund policy?

For the new subscription migration, xAI says existing subscribers can switch and receive a prorated refund for their old plan. That is a specific switching offer, not a general refund guarantee for every charge. Review the offer in your account's billing settings.

Did Cursor pricing change in 2026?

Yes. Cursor is now sold through The X Subscription for new customers. Its included Grok pool is shared across Cursor and Grok and refreshes weekly. Premium and above support third-party models at API list prices through separate top-ups; existing subscribers can keep their plan, price, and features.

What is cheaper than Cursor?

That depends on the models you use. Compare Cursor's subscription plus third-party top-ups with GitHub Copilot's included coding credits, Claude's shared subscription allowance, or Devin's included agent allowance. For a Grok-only workflow, compare against Plus rather than assuming Premium is the minimum paid tier.

Before adding another developer subscription, get the AI Business Workflow Audit Checklist.