Originally published on DevToolHub.
Cursor's pricing page lists four plans — Hobby, Pro, Pro+, and Ultra — but only one of them has a real number attached to what you get. Pro costs $20 a month for "extended limits on Agent." Pro+ costs $60 for "3x Pro limits." Ultra costs $200 for "20x Pro limits." Every tier above Hobby is defined as a multiple of Pro, and Pro's actual baseline isn't published anywhere on the page.
That's not a documentation gap. It's the whole pricing model.
What Each Cursor Plan Actually Includes
| Plan | Price | What you get |
|---|---|---|
| Hobby | Free | Limited Agent requests, access to Composer |
| Pro | $20/mo | Extended limits on Agent, frontier model access, cloud agents |
| Pro+ | $60/mo | 3x Pro's Agent limit, higher Grok Bot usage |
| Ultra | $200/mo | 20x Pro's Agent limit, priority access to new features |
| Teams | $40–$120/seat | Pooled usage across the team |
Cursor's own pricing page describes every tier's allowance in relative terms without ever stating what one unit of usage costs. A reader on justinmckelvey.com summed it up: "Every tier is a multiple of Pro, and Pro is a multiple of nothing you can see."
The Multiplier Pricing Actually Has a Pattern
Compare tiers by dollar-per-unit instead of the marketing copy and something real shows up. Pro+ costs 3x Pro's price for 3x Pro's allowance — same $20 per unit, no discount for spending more. Ultra breaks that pattern: $200 for 20x Pro's allowance works out to $10 per unit, half the per-unit cost of Pro or Pro+.
Ultra is the only tier that actually changes the underlying economics. Pro+ just buys more of the same rate.
Cursor Router Changed What "Auto" Actually Bills
Auto mode used to mean a flat rate no matter which model handled your request. That changed July 22, 2026, when Cursor Router replaced Auto's fixed heuristics with a classifier trained on 600,000+ real coding requests, picking a model per request across three modes:
- Intelligence — frontier-model quality
- Balance — strong quality at lower cost, the default for most people
- Cost — cheapest mode clearing a quality bar
Cursor's numbers show Auto Intelligence at roughly 60% lower cost than running frontier models directly, and Balance at about 36% cheaper, with measured cost-per-commit of $6.76 (Intelligence) and $4.63 (Balance).
The catch is billing, not quality. A user on Cursor's forum put it directly: "Auto, to me, used to mean 'cheap flat rate, don't worry which model actually ran behind the scenes.' I emphasized used to because as of today, this no longer holds." Auto Balance can route to an expensive frontier model and bill at its real rate — no separate warning shown.
What Happens When You Hit the Limit
Cursor tracks two usage pools: Cursor Models (Grok 4.5–4.7, Composer 2.5) and Other Models (third-party, billed at provider rates). Usage resets monthly and doesn't roll over.
Hit your limit and you get an in-editor notification with two choices: turn on on-demand pay-as-you-go usage, or upgrade tiers. Neither happens automatically. One widely shared example documented a team whose $7,000 annual subscription was consumed in a single day of normal usage.
Important: on-demand usage has no built-in spending cap of its own — check it periodically if you enable it.
Teams, Enterprise, and Legacy Max Mode Pricing
Teams/Enterprise plans using their own API key pay a flat Cursor Token Rate of $0.25 per million tokens. Legacy Max Mode bills at the model's own API rate plus a 20% surcharge. Neither rate appears on the main pricing page — both are buried in the usage-limits docs.
Quick Summary:
- Pro ($20), Pro+ ($60, 3x Pro), Ultra ($200, 20x Pro) are priced as multipliers of an unpublished baseline — only Ultra lowers the per-unit cost
- Cursor Router (July 22, 2026) replaced fixed Auto-mode routing with per-request classification across Intelligence/Balance/Cost
- Auto Balance can bill frontier-model rates with no warning — check your invoice for unexpected model names
- Teams/Enterprise with their own key pay $0.25/million tokens; legacy Max Mode is API rate + 20%
- Hitting your limit doesn't auto-charge — you choose on-demand usage or an upgrade
Top comments (0)