Cursor Pricing Explained (2026): Plans, Usage Limits, and What You Actually Pay

Cursor pricing now runs from free to $200 a month under a new owner. On October 13, 2026, the company now known as SpaceX Super Intelligence, which completed its $60 billion all-stock acquisition of Cursor on August 14, folded Cursor into The X Subscription, one bundle for Grok, Cursor, and X 12. The new ladder is Free at $0, Plus at $8, Premium at $30, Super at $100, and Ultra at $200 for individuals; Teams seats at $30, $100, or $200 per user; and Enterprise at custom pricing 34. Every paid plan bundles a weekly pool of Cursor-model usage shared with Grok, and when that pool empties, prepaid credit keeps you going at the same per-token rates, with third-party models billed at API list price and no markup 4. The ladder this page has tracked all year, Hobby, the India-only Start at ₹649, Pro at $20, Pro+ at $60, and the $200 Ultra, survives as legacy plans that keep working for existing subscribers 5. One figure still frames the spend: Cursor's docs, at the September 2026 check, put daily Agent users at $60 to $100 a month in total usage, which meant a $20 Pro seat covered only a fifth to a third of the bill. On the new ladder that job maps to the $100 Super tier, Cursor's recommendation for daily agent users 43.

Every Cursor plan, from the new Free tier to custom Enterprise

New customers buy The X Subscription:

• Free: $0 a month, limited Agent, Chat, and Tab usage on Cursor's own models 3
• Plus: $8 a month, base usage, Cursor's own models at fixed settings, Grok Chat, Voice, Imagine, Build, and Grok API access, no third-party models 34
• Premium: $30 a month, 4x Plus usage, adding Cloud Agents and Automations, code hosting on Origin, code review with Bugbot, Grok Bot, and third-party models at API list price 34
• Super: $100 a month, 15x Plus usage, priority model access at peak times 34
• Ultra: $200 a month, 50x Plus usage, earliest access to new products and models 34
• Teams: $30, $100, or $200 per user per month, matching the Premium, Super, and Ultra tiers, with seat types mixed on one team 45
• Enterprise: custom pricing, with pooled usage across Grok and Cursor, invoice and PO billing, SCIM seat management, audit logs, granular spend controls, and repository, model, and MCP access controls 35

Existing subscribers keep the legacy ladder, closed to new customers, with Cursor's own mapping to the new tiers:

• Hobby: free, maps to Free 5
• Start (India only): ₹649 a month, tax inclusive, maps to Plus 5
• Pro: $20 a month, maps to Premium 5
• Pro+: $60 a month, maps to Super 5
• Ultra: $200 a month, maps to Ultra 5
• Teams Standard: $40 per user per month, maps to a Premium seat; Teams Premium: $120 per user per month, maps to a Super seat 5

Billing on the new ladder is monthly or yearly, and annual pricing buys twelve months for the price of ten: $80 a year for Plus, $300 for Premium, $1,000 for Super, and $2,000 for Ultra, all excluding tax 45. The legacy ladder carried a 20 percent annual-discount note at the September 2026 check, when Cursor's pricing page printed "Save 20% with yearly billing", and Vantage's guide, updated March 4, 2026, listed the same 20 percent 36. Neither the pricing page nor the help pages this desk could read print a legacy annual price at the October 16 check. Switching from a legacy plan is optional and one-way: the legacy subscription is cancelled with a prorated refund, the pick on The X Subscription is permanent, and legacy subscribers can still move between legacy plans, Pro to Pro+ for example, without switching 5.

One weekly pool on the new plans, two monthly pools on legacy

The X Subscription gives every paid plan one pool of Cursor-model usage, and the pool refreshes weekly, so one heavy day does not use up the month 4. It is shared across everything on the subscription: Agent in the editor, the CLI, Cloud Agents, Grok Bot, and the Grok apps, with Grok Chat, Imagine, Voice, and Bot all drawing from the same pool at product-specific rates 43. The Grok API is the exception: paid plans include access, but API usage bills separately at standard model rates and does not draw from the pool 3. The pool covers Cursor's own models, Grok 4.7, Grok 4.6, Grok 4.5, and Composer 2.5, and different models draw from it at different rates; remaining usage and the next reset date show in editor settings and at the usage dashboard 4. Plus runs those models at fixed settings, Fast off, medium effort, Grok 4.7 at the 256k context window; Premium and above unlock Fast mode and effort choices plus third-party models at API list price paid from prepaid credit 4. On Cursor's plan pages the pools scale by multiplier against Plus: 4x on Premium, 15x on Super, 50x on Ultra 34. The legacy ladder keeps its own structure: two monthly pools, Cursor Models and Other Models, with on-demand usage for extra requests instead of top-ups 45. Vantage's March 2026 guide pegged those legacy pools in dollars: about $20 of API usage on Pro, $70 on Pro+, $400 on Ultra, and $20 per Teams seat 6.

Auto, the mode that picks a model for each request, now runs through Cursor Router in three modes: Cost, Balance, and Intelligence 45. Cost mode, the classic Auto, bills as a Cursor-model request and draws from the weekly pool whatever model serves it, and it is the only mode on individual plans 4. Balance and Intelligence are for Teams and Enterprise with Cursor Router switched on: a route to a Cursor model draws from the pool, and a route to a third-party model bills at that model's list price as third-party usage paid from prepaid credit, plus the Cursor Token Rate 45. Router launches for Teams and Enterprise first, and Enterprise teams start with it off until an admin opts in 5. One dated warning for anyone comparing older guides: Vantage's March 2026 guide prices Auto at "roughly $0.25/M tokens (cache read), $1.25/M tokens (input), and $6.00/M tokens (output)" 6. Those exact rates only ever matched Cursor's Legacy Enterprise Auto schedule, a flat price regardless of model, which ran until September 7, 2026, and no longer appears anywhere in the docs this page could read 4.

What a million tokens costs once the pool runs out

On the new plans, usage past the weekly pool bills from prepaid credit, bought as top-ups or refilled by Auto Top-up, and it bills at the same per-model rates as the pool; a monthly additional-usage limit caps paid usage regardless of how much credit you hold, and Cursor states that requests are never downgraded in quality or speed when limits hit 43. On legacy plans the same rates bill as on-demand usage, monthly and in arrears 4. Third-party models are not part of any pool: from the first request they bill at the model's API list price out of prepaid credit, on Premium and above only 4. The lever that moves the bill is model choice, not an overage penalty. The current rate card for the 20 model entries Cursor's docs list up front, in USD per million tokens, input then output, with more folded behind a Show more models link 4:

• Cursor Models pool: Composer 2.5 at $0.50 input and $2.50 output; Composer 2.5 Fast at $3 and $15; Grok 4.5 at $2 and $6; Grok 4.5 Fast at $4 and $18; Grok 4.6 at $2 and $6; Grok 4.6 Fast at $4 and $12; Grok 4.7 at $2 and $6; Grok 4.7 Fast at $4 and $12; Grok 4.7 500k at $4 and $12; Grok 4.7 500k Fast at $6 and $18
• Third-party models at list price: Claude Fable 5.1 at $10 and $50; Claude Haiku 5.5 at $0.10 and $0.50; Claude Opus 5.5 at $4 and $20; Claude Sonnet 5.5 at $2 and $10; Gemini 3.1 Pro at $2 and $12; Gemini 3.8 Flash at $0.75 and $3.75; Gemini 4 Argon at $2 and $10; GPT-5.6 Luna at $0.20 and $1.20; GPT-5.6 Sol at $4 and $20; GPT-5.6 Terra at $2 and $12; Muse Wonder at $4 and $20

Three entries moved between the October 2 and October 16 checks: Gemini 3.8 Flash output rose from $3.50 to $3.75, Muse Spark 1.3, at $1.25 input and $4.25 output, left the card, and Muse Wonder, Claude Haiku 5.5, and Gemini 4 Argon joined it 4. Cache pricing cuts the sting: cache reads bill as low as a twentieth of fresh input, $0.10 per million tokens on Claude Sonnet 5.5 for example, so workloads that reuse the same context cost less than raw token counts imply 4. Opting into regional data residency adds a 10 percent uplift on eligible models 4. On legacy request-based plans, Max Mode still extends a model's context window at the model's API rate plus 20 percent 4.

Same workload, 20x the bill: the arithmetic of model choice

Price one illustrative heavy month for a single developer, 100 million input tokens and 10 million output tokens, ignoring cache discounts, across four of the rates above 4:

• Composer 2.5 at $0.50 input and $2.50 output per million: $50 input plus $25 output is $75
• Grok 4.6 at $2 input and $6 output per million: $200 input plus $60 output is $260
• Claude Sonnet 5.5 at $2 input and $10 output per million: $200 input plus $100 output is $300
• Claude Fable 5.1 at $10 input and $50 output per million: $1,000 input plus $500 output is $1,500

Identical tokens, $75 on Composer 2.5 versus $1,500 on Claude Fable 5.1: a 20-to-1 spread with no change of plan. The $75 result lands inside the $60 to $100 band Cursor's docs published for daily Agent users at the September 2026 check 4. The identical workload on Claude Fable 5.1 runs 15 to 25 times that entire monthly band. That spread, not an overage markup, is the mechanism behind the unpredictable-bill complaints that followed Cursor's June 2025 move from request-based pricing to token-based billing 76.

What Cursor recommends, and what a daily Agent user actually pays

Cursor's current pages recommend by fit rather than dollars: Premium for new AI users, Super for daily agent users, Ultra for power users, and Plus for social creators with limited AI usage 34. The last published dollar bands, on the docs at the September 2026 check, read: daily Tab users typically stay within included usage, limited Agent users often stay within included usage, daily Agent users typically $60 to $100 a month in total usage, and power users running multiple agents or automations often $200 or more 4. No dollar expectation band prints anywhere on the pages this desk could read at the October 16 re-check; the tier recommendations are all that remain 345. The arithmetic still explains the ladder. A daily Agent user on legacy Pro paid $20 for the seat and $40 to $80 on top in on-demand charges, so the subscription was 20 to 33 percent of the bill, and the same workload now maps to the $100 Super seat. When included usage runs out on the new plans, the docs offer two moves: prepaid credit at the same rates, or an upgrade, which takes effect immediately, while downgrades schedule for the end of the billing period 45. Legacy upgrades run through the Cursor billing dashboard with credit based on unused included usage rather than days remaining, and legacy downgrades schedule the same way 5. On Teams, admins manage a shared prepaid balance and one cap for the whole team 4.

The ₹649 Start plan: the cheapest paid seat, now behind a legacy fence

Start remains Cursor's plan for developers in India, launched July 27, 2026 in what TechCrunch called Cursor's "biggest India push yet" 8. It costs ₹649 a month, tax inclusive, the only Cursor plan priced after tax, billed monthly in rupees 4, which works out to about $7, roughly a third of Pro's $20 9. Since the October 13 bundle it is closed to new customers and maps to the $8 Plus tier for anyone who switches 5. The fences, as the docs listed them at the October 2 check, still bind whoever holds the plan: no Other Models pool, no Bugbot, no Auto, no Automations, and Grok 4.7, Grok 4.6, and Grok 4.5 pinned to a fixed medium effort level with Fast mode unavailable 4. Plus, its new-ladder counterpart, carries the same fixed-settings treatment: Fast off, effort medium, no third-party models, and no own API keys 4. For an India-based developer working in Cursor's own models, Start still undercuts every paid seat on either ladder. The first need for a Claude or GPT model, or for usage past the pool, now sends the upgrade path to Premium or above 45.

Why a Teams bill runs higher than its sticker price

Two adders stack onto Teams seats. The first is the Cursor Token Rate: on Teams and Enterprise, third-party model requests carry an extra $0.25 per million tokens on top of the model's API price, applied when you pick the model directly and when Cursor Router routes to it; Cost-mode requests and Cursor's own models, including Grok and Composer, are exempt, and individual plans do not pay it 4. On the 100-million-input, 10-million-output workload above, that is $27.50. On GPT-5.6 Luna, the cheapest third-party model at $0.20 input and $1.20 output per million tokens, the workload costs $32 in tokens and $27.50 in Token Rate, an 86 percent surcharge on the model's own cost 4. New-ladder Teams seats price at $30, $100, or $200 to match the Premium, Super, and Ultra tiers, with each member getting the same weekly usage as the matching individual plan and Plus unavailable as a team seat; legacy Teams Standard at $40 and Teams Premium at $120 continue for existing teams 45.

The second adder is the one you cannot see. Since July 31, 2026, Cursor's usage dashboard, CSV exports, and API have returned token counts instead of dollar figures for self-serve plans including Teams, and the change zeroed historical dollar data retroactively; a Cursor staff member, Kevin Neilson, confirmed the removal was deliberate 7. Enterprise plans kept full dollar visibility, while Teams admins can still see a per-user on-demand total with no per-model breakdown; one Teams administrator reported a $30,000 billing cycle while asking how to track spend by model and by user 7. For a Teams buyer, the practical workaround is to price the dashboard's token counts against the rates on this page. The removal landed just over a year into metered billing: Cursor moved Pro from a fixed 500-request monthly limit to token-based pricing in June 2025, then apologized on July 4, 2025 and refunded surprise charges incurred between June 16 and July 4 7.

The cheapest path for each kind of Cursor user

• Trying it out: Free, at $0 3
• Social creators and light AI use: Plus at $8, Cursor's own models at fixed settings, Grok Chat, Voice, Imagine, and Build included 34
• New to Agent work: Premium at $30, Cursor's recommendation for new AI users, the first tier with third-party models, Cloud Agents, Bugbot, and Origin 34
• Daily Agent work: Super at $100, Cursor's recommendation for daily agent users 34
• Multi-agent, automation-heavy use: Ultra at $200, matching the $200-or-more band the docs published for power users at the September check 4
• Teams: Premium, Super, or Ultra seats at $30, $100, or $200 per user, mixed as needed 4
• On a legacy plan: keep it, or switch once with a prorated refund; the pick is permanent 5
• Anyone who needs per-model dollar reporting: Enterprise, the tier that still shows it 37

The model lever still beats the plan lever. Moving the illustrative workload from Claude Fable 5.1 to Composer 2.5 saves $1,425 in a month, within $15 of what four years of the $30 Premium plan costs. Before upgrading a seat, check what the seat is routing to.

Verified October 2026: prices, tiers, pools, and per-token rates re-checked against Cursor's pricing, docs, and help pages on October 16, 2026, three days after The X Subscription launch, with the launch itself checked against TechCrunch's October 13 reporting.