Cursor promises more coding usage per dollar through The X Subscription. But switching economics depend on which models customers actually use. Our public-document comparison finds two light Claude or GPT workloads cheaper after switching a Teams Standard seat—even with separate model bills. For individual Pro customers using only third-party models, switching to X Premium instead costs at least $10 more monthly.
The distinction sits inside Cursor’s October 12 announcement. The company says comparable individual tiers deliver at least twice as much coding usage with Grok, and team tiers provide three times as much. Those are company claims about Grok usage, not findings that every customer’s total bill falls. Claude and GPT remain available on Premium and higher, but their usage is billed separately.
Existing Cursor plans include separate monthly pools for first-party models, such as Grok and Composer, and third-party models. The new subscription instead includes a shared weekly pool for eligible first-party usage across Grok and Cursor surfaces. Third-party models draw on purchased additional-usage credits from the first request. Switching therefore changes both the seat price and what that price buys.
The comparison holds model demand constant
This Original analysis synthesizes supplied pricing, usage, migration and official staff records; it is not a product test or an audit of customer invoices. We compare existing US individual Pro, Pro Plus and Ultra plans with selected X tiers, alongside self-serve Teams Standard and Premium seats. The calculations use monthly dollar list prices and identical, directly selected hypothetical model workloads. Taxes, regional data-residency charges, ancillary paid features and separate Grok or X subscriptions are excluded.
Our requested evidence cutoff was October 13, 2026, at 02:31:47.468 UTC. Supplied snapshots were retrieved at 02:33:00.270 UTC, afterward. We use these records to compare documented terms, not to certify exact availability at that earlier instant. The selected plan pairings are price comparisons, not verified mappings of equivalent tiers. Premium or higher is used for third-party scenarios because that is where the launch announcement confirms access.
A $10 seat saving absorbs a small model bill
Teams Standard costs $40 per user monthly; X Premium costs $30 per seat. Consider a monthly workload with one million uncached input tokens and 200,000 output tokens, with no other usage. Tokens are the billable units of material sent to a model and generated by it. At Cursor’s listed Sonnet 5.5 rates, that workload costs $4 before the team surcharge. GPT-5.6 Sol costs $8.
The $0.25-per-million team token charge adds $0.30 in either case. That produces X Premium totals of $34.30 for Sonnet or $38.30 for Sol. Both fall below the existing $40 subscription alone, so these two cases save money regardless of the old seat’s usable third-party allowance. Any old-plan overages would only increase the old total.
A $30 seat plus $4 model usage and $0.30 surcharge; $5.70 below the old subscription alone.
A $30 seat plus $8 model usage and $0.30 surcharge; $1.70 below the old subscription alone.
The surcharge is not a new switching penalty. Existing Cursor Teams already applies the same rate to third-party input, output and cached tokens, including included usage. It survives the migration rather than appearing because of it. Individuals do not pay that charge. Grok and Composer are exempt, but Composer’s exemption alone does not establish that its usage is included in the new weekly pool.
The individual result runs the other way
Selected switching outcomes
- Costlier: Pro at $20 to X Premium at $30 adds at least $10 monthly for third-party-only use, before counting any lost included credits.
- Costlier when credits are used: Ultra to X Ultra keeps the $200 subscription price but replaces included third-party usage with separate charges.
- Conditional: Ultra to X Premium crosses from cheaper to costlier at $170 of monthly third-party API spend, if the August-documented $400 allowance still applies.
- Publicly indeterminate: A specified Grok workload cannot be fully priced against both plans without usable allowance quantities and the relevant overage inputs.
The Ultra crossover relies on dated evidence. On August 10, Cursor staff member Kevin Neilson identified $400 of included third-party usage for Ultra. That is not an October account-level guarantee. If it remains applicable, $100 of hypothetical third-party spend means $200 on old Ultra versus $130 on X Premium. At $200 of spend, the comparison becomes $200 versus $230.
The general rule is straightforward under dollar-credit accounting: switching adds the subscription-price difference, plus the old third-party credits actually used, plus any change in first-party overages. A cheaper subscription wins only when that saving exceeds the other two costs. For teams without first-party overages, Standard-to-Premium starts with a $10 saving; Premium-to-Super starts with $20. Utilized legacy allowance erodes those savings.
Weekly resets can change the value of a pool
For Grok-only workloads that fit both plans’ allowances in every relevant period, subscription prices settle the comparison. Pro-to-X-Premium costs $10 more; Pro-Plus-to-X-Super costs $40 more; Ultra-to-X-Ultra is equal. The selected team pairings save $10 or $20 per seat. But those results assume the workload fits. Public documentation does not supply enough quota detail to verify that assumption for a specified Grok token workload.
Timing also matters. In a hypothetical four-window example, consuming one weekly quota in each window fits. Consuming four quotas in the first window exceeds that window’s allowance by three quotas and leaves later capacity unused. This assumes no rollover, which the reviewed text does not explicitly specify. It illustrates possible lost utilization, not a dollar overage or a conversion of four weeks into a calendar month.
Other eligible Grok-product activity draws from the same seat’s pool, and product consumption rates vary. On teams, idle colleagues cannot donate their weekly allowances; only the prepaid additional-usage balance is shared. These rules affect first-party capacity, not whether Claude or GPT tokens are included. Moving third-party requests between weeks does not turn them into bundled usage.
Reader comments
Newest comments first. Replies stay oldest first.