Cursor and The X Subscription: how third-party models are priced
As of 2026-10-13 the Cursor help page defines it word by word: The X Subscription is one subscription that brings Grok, Cursor, and X together, followed by Grok and Cursor share a weekly included usage pool. The same page states that third-party models in Cursor, like Claude or GPT, aren't included in your plan and are billed at API list prices from the first use against your prepaid credit. cursor.com/pricing the same day also says Premium and above do not include an allowance for third-party models.
Updated 2026-10-13
One key across Claude, GPT and Chinese models, billed per token.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard / AMEX), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Four pricing points to check first
Cursor list prices (2026-10-13)
Plus, Premium, Super and Ultra as listed on cursor.com/pricing; the footer says Prices subject to change and the stated prices exclude third-party model usage.
The billing basis for other models
Cursor's own line is Third-party models remain available in Cursor at API list price, and the footnote adds that such usage must be purchased separately.
Cursor Token Rate
Cursor states that Teams and Enterprise requests to third-party models carry $0.25 per million tokens, applying to included, on-demand and BYOK usage, with first-party Grok and Composer exempt.
Two refresh wordings coexist
The X Subscription overview and the pricing FAQ say the allowance refreshes weekly, while docs/models and usage-limits describe two pools resetting with the monthly billing cycle. Both were live on 2026-10-13.
What the X Subscription has to do with a Cursor plan
As of 2026-10-13 the definition on Cursor's help page is The X Subscription is one subscription that brings Grok, Cursor, and X together, immediately followed by Grok and Cursor share a weekly included usage pool, meaning Grok and Cursor draw on the same included allowance. Switching is optional: the same page says Switching to The X Subscription is optional and Your current plan, price, and features continue unless you change or cancel, and that existing Cursor customers can still upgrade or downgrade between their current Cursor plans without switching. Billing sits at workspace level - the page says a workspace can be for an individual or a team - with the subscription, bill, usage and top-ups managed there. After a switch, Cursor writes that eligible Cursor and Grok accounts are combined under one subscription, workspace, and bill. Third-party models are listed separately: third-party models in Cursor, like Claude or GPT, aren't included in your plan, remain usable, and are billed at API list prices from the first use and paid from your prepaid credit.
What the official pages say on 2026-10-13
Every Cursor-side sentence here comes from cursor.com pages checked on 2026-10-13: the pricing page, docs/models-and-pricing, help/the-x-subscription/overview, the api-keys, usage-limits and token-rate help pages, and help/account-and-billing/pricing. On that date neither cursor.com/blog nor cursor.com/changelog carried an X Subscription announcement - the newest changelog entry is Oct 6, 2026, Remote control for local agents - and Cursor has not published a launch date for the X Subscription, so the wording quoted here is what those pages said that day.
The timeline the official pages support
The joining-spacex post on cursor.com/blog is dated Aug 14, 2026 and states Cursor has officially been acquired by SpaceX, adding that this completes the acquisition process that started in April.
As of 2026-10-13 the newest entry on cursor.com/changelog is Oct 6, 2026, Remote control for local agents.
The same day cursor.com/pricing still lists Plus $8, Premium $30, Super $100 and Ultra $200, and both the pricing page and the X Subscription overview state that third-party models are billed separately at API list prices.
Confirmed vs not published
Confirmed (the wording on cursor.com on 2026-10-13)
Each of these can be found word for word on the cursor.com pages as of 2026-10-13: the X Subscription definition sentence and the weekly shared pool for Grok and Cursor; switching being optional, with current plans, prices and features continuing and existing customers still able to move between current Cursor plans without switching; workspaces covering an individual or a team, holding the subscription, bill, usage and top-ups; third-party models excluded from the plan and billed at API list prices from first use against prepaid credit; the pricing footnote that third-party model usage is not included in the monthly prices shown and must be purchased separately; Plus includes Grok models only; Premium and above carrying no allowance for third-party models; the $0.25 per million token Cursor Token Rate on Teams and Enterprise, charged both when you pick a third-party model and when Auto routes to one, with Grok and Composer exempt; Auto modes billing at the routed model's list price; and subscriptions sold only directly by X Subscription Corp, with no resellers authorized.
Unverified and not published by the vendor
On the cursor.com pages quoted here as of 2026-10-13 - pricing, docs/models-and-pricing, help/the-x-subscription/overview, the api-keys, usage-limits and token-rate pages, and help/account-and-billing/pricing - there are no allowance figures for third-party models on the unified tiers, no consolidated X Subscription price list and no regional tax-inclusive amounts; the overview page links to docs.x.ai for detail, and the quotations here stay with the cursor.com pages named above. One wording conflict is worth flagging: cursor.com/pricing uses Plus, Premium, Super and Ultra and says the allowance refreshes weekly, while help/account-and-billing/pricing names the plans Hobby, Pro, Pro+ and Ultra, and docs/models lists Pro $20/mo, Pro Plus and Ultra $200/mo with two pools resetting monthly, and the help table also lists Start (India only) at ₹649/mo, tax inclusive, plus Teams Standard and Teams Premium at $40/user/mo and $120/user/mo. Both namings were on the official pages on 2026-10-13, and the official wording of that day governs. Posts and view counts on X are not official statements.
Two payment paths, sorted by what you actually use
Staying inside the included allowance
Cursor's FAQ says Every tier includes a pool of AI usage shared across Cursor and Grok and that your allowance refreshes weekly; Plus is documented as Plus includes Grok models only. On this path you pay the monthly subscription price (listed 2026-10-13 as Plus $8, Premium $30, Super $100, Ultra $200); third-party models are billed separately, and the same holds when Auto routes a request to one: All Auto modes bill at the list price of the model each request is routed to.
Using third-party models or your own key
Cursor states Premium and above include generous usage of Grok models but do not include an allowance for third-party models, and that third-party models remain available in Cursor at API list price; the overview adds they are billed from the first use against prepaid credit. With your own key Cursor says your provider bills you directly for the model cost and, on individual plans, those requests do not draw from either pool - but custom API keys only work with chat models. Teams and Enterprise still add $0.25 per million tokens. These are Cursor's own terms and say nothing about how QCode bills.
Three checks before you read your bill
① Confirm whether your tier carries a third-party model allowance at all: the pricing page says Plus includes Grok models only, that Premium and above do not include an allowance for third-party models, and that such models are billed at API list price separately from the monthly prices shown. ② Decide about your own key: the key goes in Cursor Settings under Models, and Cursor says your provider bills you directly for the model cost of those requests, and custom API keys only work with chat models while Tab completion keeps using built-in models. ③ Add the Token Rate if you are on Teams or Enterprise: Cursor writes that third-party model requests include a Cursor Token Rate of $0.25 per million tokens, on top of model API pricing, for included, on-demand and BYOK usage alike. For Auto, Cursor's own example reads if your request is routed to Opus 5, you are billed at Opus 5 pricing for that request.
Using QCode inside Cursor
Cursor reaches QCode over the OpenAI-compatible endpoint. The QCode page for Cursor puts the address under Settings, then Models, then API Keys, the OpenAI slot takes https://api.qcode.cc/openai/v1, and it warns against a trailing slash on the Base URL; a 401 from the built-in check means the path is right and only authentication is missing. Create a key starting with cr_ in the console, then tick the model IDs you want in the Models list, for example glm-5.3 or glm-5.3-flash; the live list of IDs and rates is on /models. Which IDs Cursor lets you tick is Cursor's own call: its help page limits custom OpenAI keys to Standard, non-reasoning chat models and states The model picker shows which OpenAI models are available, so reasoning models may not be accepted inside Cursor - check each one there. For Claude, use a tool that speaks the Anthropic protocol, or Claude Code, on https://api.qcode.cc/api. Billed per token; per-model rates are listed on /models. The Cursor plan prices, API list prices and Token Rate on this page are Cursor's numbers and do not apply to QCode.
Frequently asked questions
What is the relationship between the X Subscription and a Cursor plan?
One subscription covering Grok, Cursor and X. The help page states The X Subscription is one subscription that brings Grok, Cursor, and X together and then Grok and Cursor share a weekly included usage pool. Moving over is not forced: the same page says Switching to The X Subscription is optional and Your current plan, price, and features continue unless you change or cancel. After a switch Cursor says eligible Cursor and Grok accounts are combined under one subscription, workspace, and bill.
How are third-party models priced inside Cursor?
Outside the plan, at the model's own API list price. cursor.com/pricing on 2026-10-13 reads Third-party models remain available in Cursor at API list price and Premium and above do not include an allowance for third-party models; the footnote adds that such usage is not included in the monthly prices shown and must be purchased separately. The X Subscription overview says they are billed at API list prices from the first use and paid from your prepaid credit, and docs/models states All prices are per million tokens.
Does my own API key consume included usage, and what is the Cursor Token Rate?
On individual plans it does not: Cursor writes On individual plans, no. Your provider bills you directly for the model cost, and those requests don't draw from either pool. The Cursor Token Rate is $0.25 per million tokens charged on third-party model requests for Teams and Enterprise customers; Cursor states it applies on top of model API pricing to included, on-demand and BYOK usage, with first-party Grok and Composer exempt. Two further limits: custom API keys only work with chat models, and every request still passes through Cursor's servers for final prompt building.
How do I pay and get started from mainland China?
You can top up with Alipay or WeChat Pay on the CNY side. QCode is prepaid: you top up, then buy a plan from the balance, and you create a cr_ key in the console and paste it into your tool. The payment overview also lists Visa / Mastercard / AMEX plus Apple Pay and Google Pay on the USD side, and cryptocurrency across many coins and networks. Billed per token; per-model rates are listed on /models. Paying for a Cursor plan is a separate transaction between you and Cursor.
How do I point Cursor at a GPT or domestic model through QCode?
In Cursor open Settings, then Models and the API Keys field, and put https://api.qcode.cc/openai/v1 in the OpenAI slot with no trailing slash, paste the cr_ key created in the console, then tick the model IDs you want in the Models list, for example glm-5.3 or glm-5.3-flash - what you can tick is whatever Cursor's picker shows. Cursor's help page limits custom OpenAI keys to Standard, non-reasoning chat models, so reasoning models may not be accepted there. Billed per token; per-model rates are listed on /models. Claude is not on that path: use a tool that supports the Anthropic protocol, or Claude Code.
How is Auto mode billed?
At the list price of whichever model the request is routed to. Cursor states All Auto modes bill at the list price of the model each request is routed to, and its pricing help page gives the example if your request is routed to Opus 5, you are billed at Opus 5 pricing for that request. Cursor adds that third-party models also incur the Cursor Token Rate and that, since models have different API costs, your model selection affects how quickly your included usage is consumed.
Sources
Cursor quotations come from cursor.com pages checked on 2026-10-13: the pricing page, docs/models-and-pricing, help/the-x-subscription/overview, the api-keys, usage-limits and token-rate help pages, help/account-and-billing/pricing, blog/joining-spacex and the changelog listing. QCode statements come from the docs.qcode.cc pages on Cursor setup, endpoints and payment methods from the same date, plus the matching qcode.cc guide. Quoted sentences stay in their original language.
A per-token route for your editor
Put https://api.qcode.cc/openai/v1 in Cursor's OpenAI slot and tick model IDs such as glm-5.3 or glm-5.3-flash where Cursor accepts them. Billed per token; per-model rates are listed on /models.
Related reading
Cursor Custom API Key & Base URL: OpenAI Override (2026)
Where Cursor takes your own key, why the Base URL override is documented for the OpenAI slot, and that keys cover chat models.
Cursor Auto pricing change: what to actually check
When Auto pricing moves, what shifts your cost is which requests count toward usage; that page is the checklist.
Claude Code vs Cursor: In-Depth Comparison of Two Leading AI Coding Tools
How the two tools divide work across models, context, execution mode and agent capability.
Cursor figures and quotations on this page were checked on 2026-10-13 against that day's cursor.com pages; official pages change at any time, so the official page governs. The QCode model list and rates are governed by /models. Cursor plan prices, API list prices and the Cursor Token Rate are Cursor's own terms and are not QCode prices or commitments.