Cursor Pricing: What the Plans Actually Give You

cursor pricing breakdown showing Hobby, Pro, Pro+, and Ultra plan tiers

Cursor pricing lists four plans — Hobby, Pro, Pro+, and Ultra. Only one number on the page is real: the price. 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. Pro’s actual baseline — the request count, token count, or dollar value it represents — isn’t published anywhere on the page.

That’s not a documentation gap. It’s the whole pricing model. It’s also why so many people search for what these plans actually include before they pay for one.

What Cursor Pricing Actually Includes at Each Tier

Cursor’s four consumer plans are Hobby (free), Pro ($20/month), Pro+ ($60/month), and Ultra ($200/month). Teams costs $40 or $120 per seat, and Enterprise is priced on request. Annual billing saves 20% across every paid tier.

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

Notice what’s missing from that table: an actual number. Cursor’s own pricing page describes every tier’s allowance in relative terms — “extended,” “3x,” “20x.” It never states what one unit of Agent usage costs, or how many you get for $20. A reader on justinmckelvey.com put it plainly after going through the page line by line: “Every tier is a multiple of Pro, and Pro is a multiple of nothing you can see.”

Why Cursor Pricing Multipliers Actually Have a Pattern

Compare tiers by dollar-per-unit instead of the marketing copy. The pricing does reveal something real. Pro+ costs 3x Pro’s price for 3x Pro’s allowance. That’s the 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+.

So the only tier that changes the underlying economics is Ultra. Already paying for Pro+ and hitting limits regularly? Ultra is a better dollar-for-dollar deal, not just a bigger number. Pro+ itself doesn’t buy a discount over Pro. It buys more of the same rate. The GitHub Copilot AI Credits breakdown walks through the same “read past the label” math for Copilot’s own plans.

Cursor Router Changed What “Auto” Actually Bills

Auto mode used to mean a flat, predictable rate no matter which model handled your request. That changed on July 22, 2026. Cursor replaced Auto’s routing logic with Cursor Router, a classifier trained on more than 600,000 real coding requests. It picks a model per request instead of using fixed heuristics.

Router runs in three modes. Here’s the breakdown:

Intelligence — frontier-model quality, matching the most capable models available
Balance — strong quality at lower cost, the default most people run day to day
Cost — the cheapest mode that still clears a quality bar

Cursor’s own numbers show Auto Intelligence landing near frontier-model output. The cost is roughly 60% lower for teams. Auto Balance runs about 36% cheaper than Opus 4.8 in the same tests. The measured cost-per-commit is $6.76 for Intelligence mode and $4.63 for Balance.

The catch is billing behavior, not model quality. A user named Shuvo raised this directly on Cursor’s community forum: “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 now route a request to a genuinely expensive model. It bills that model’s real rate, with no banner warning you it happened. If your invoice shows “GPT-5.6 Sol (Auto Balanced),” that’s expected behavior under the new system, not a bug.

What Happens When Cursor Pricing Limits Kick In

Cursor tracks usage in two separate pools, according to its usage-limits documentation. A Cursor Models pool covers Grok 4.5 through 4.7 and Composer 2.5. An Other Models pool handles third-party models, billed at provider rates. Usage resets with your monthly billing cycle. Unused usage doesn’t carry over.

When you hit your limit, Cursor shows an in-editor notification. You get two options: enable on-demand usage (pay-as-you-go past your plan’s allowance), or upgrade to a higher tier. Neither happens automatically. You have to choose, which is also why surprise bills keep showing up in Cursor discussion threads. One widely shared example documented a team whose $7,000 annual subscription got consumed in a single day. They described that usage as normal.

⚠️ Important: if you turn on on-demand usage to avoid interruptions, check it periodically. It has no built-in spending cap of its own. That’s a separate setting.

Teams, Enterprise, and Legacy Max Mode Pricing

Teams and Enterprise plans that connect their own API key get billed differently. The flat Cursor Token Rate is $0.25 per million tokens, separate from the consumer usage pools. Anyone still on the older Max Mode pays the model’s own API rate plus a 20% surcharge instead.

None of these rates are visible on the main pricing page. You only find them in the usage-limits documentation, worth reading in full before you commit a team to a plan. The Ollama Cloud pricing breakdown covers a similar credit system for a different tool. That one’s for local-model hosting instead of coding agents.

Frequently Asked Questions

Q: How much does Cursor Pro actually give you in requests?
A: Cursor doesn’t publish an exact request or token count for Pro. It’s described only as “extended limits on Agent” relative to Hobby’s “limited” allowance. The practical way to find out is to run your normal workload for a billing cycle. Then watch the usage dashboard.

Q: Is Cursor Pro+ worth it over Pro?
A: Pro+ costs exactly 3x Pro’s price for 3x Pro’s allowance — the same rate per unit. It’s worth it if you consistently hit Pro’s limit before the month resets. It doesn’t improve your cost per request the way Ultra does.

Q: Why did my Cursor bill suddenly go up on Auto mode?
A: Cursor Router (launched July 22, 2026) can route Auto Balance requests to higher-cost frontier models. It bills at that model’s real rate, without a separate warning. Check which mode you’re running. Switch to Cost mode if predictable billing matters more than peak output quality.

Q: What happens if I hit my Cursor usage limit mid-month?
A: You get an in-editor notification and a choice: turn on pay-as-you-go on-demand usage, or upgrade to the next plan tier. Usage doesn’t roll over between billing cycles either way.

Understanding Cursor pricing this way — by unit economics instead of the label on the tier — is what actually tells you whether an upgrade saves money or just buys a bigger number.

Quick Summary:
– Pro ($20), Pro+ ($60, 3x Pro), and Ultra ($200, 20x Pro) are priced as multipliers of an unpublished baseline — only Ultra actually lowers the per-unit cost
– Cursor Router (July 22, 2026) replaced Auto mode’s fixed heuristics with per-request model classification across Intelligence, Balance, and Cost modes
– Auto Balance can now bill at frontier-model rates without a separate warning — check your invoice for unexpected model names
– Teams/Enterprise with their own API key pay a flat $0.25 per million tokens; legacy Max Mode is billed at API rate plus 20%
– Hitting your limit doesn’t auto-charge you — you choose between on-demand pay-as-you-go or an upgrade

If you’re deciding between Cursor and an alternative, the Cursor vs GitHub Copilot comparison covers the feature side; this is the piece to read before you pick a plan tier.

DevToolHub