Cursor's Six Pricing Plans: We Did the Math So You Don't Cry Later
Cursor Teams costs $40/seat but only $20 of that is AI credit. We broke down all six plans — Hobby through Enterprise — and the credit math nobody explains upfront.
Cursor’s pricing looks clean on the surface — six plans, one number per plan. Then you read the fine print on Teams and realize each $40 seat only includes $20 of actual AI usage. The rest buys admin controls and centralized billing. We ran the numbers across all six tiers so you can stop guessing which one won’t make you regret hitting Enter.
The Six Plans, Priced
| Plan | Cost | Included Usage | Best For |
|---|---|---|---|
| Hobby | Free | ~2,000 tab completions, ~50 slow requests/mo | Students, experimenting |
| Pro | $20/mo | $20 credit/mo (~225 Claude Sonnet OR ~550 Gemini requests) | Solo devs on a budget |
| Pro+ | $60/mo | $60 credit/mo (~675 Claude Sonnet OR ~1,650 Gemini requests) | Full-time solo users |
| Ultra | $200/mo | $200 credit/mo (~4,500 Claude Sonnet OR ~11,000 Gemini requests) | Power users, freelancers billing to clients |
| Teams | $40/user/mo | $20 credit per user + admin/SSO/billing | Organizations with 5+ developers |
| Enterprise | Custom | Custom pooling, invoicing, audit logs | 50+ seat orgs, regulated industries |
The credit system is the catch. Back in June 2025, Cursor ditched the old “500 fast requests per month” model and moved to a credit-based system tied to token consumption. Your actual mileage depends on which AI model you pick.
What June 2025 Quietly Changed
Cursor’s transition from requests to credits sounds neutral. It wasn’t.
Under the old system, Pro users got roughly 500 fast requests. Under the credit system, the equivalent is closer to 225 Claude Sonnet requests or 550 Gemini requests for the same $20. That’s because Claude Sonnet costs roughly 4x as much per request as Gemini. Pick the wrong model and you’ve burned half your monthly budget on half the feature work you expected.
The silver lining: Auto mode (Cursor’s simpler, faster completions) doesn’t touch your credits at all. Unlimited Auto, but frontier models (Claude, GPT-4, Gemini Advanced) eat your $20 allowance. Most devs mix the two, which means your effective request count sits somewhere between “plenty” and “oops.”
The Multiplier Math
Pro+ costs 3x as much as Pro ($60 vs $20) but gives 3x the credit pool. Ultra costs 10x as much as Pro+ but gives only 3.3x the credit pool. The pricing curve flattens as you go up.
Here’s the percentage math:
Pro: $20 per 225 Claude requests = $0.089 per request (or $0.036 per Gemini request)
Pro+: $60 per 675 Claude requests = $0.089 per request (same efficiency)
Ultra: $200 per 4,500 Claude requests = $0.044 per request (half the cost per request)
Ultra is the only tier where you actually get a discount per unit. But you have to burn $200 a month to access it. Unless you’re running agents 8 hours a day or billing time to clients, the math doesn’t justify the upfront spend. Pro or Pro+ for individuals. Ultra if you’re a freelancer passing the cost to a client.
Teams Plan: When $40/Seat Makes Sense
Each Teams seat costs $40 a month, with $20 of included AI credit. The other $20 buys SSO, admin dashboards, usage analytics, and centralized billing. You’re not paying twice for the AI — you’re paying once for AI and once for team features.
This means:
Five people on Teams: $200/month total, $100 in combined AI credits
Five people on individual Pro subs: $100/month total, $100 in combined AI credits
Five people on individual Pro+ subs: $300/month total, $300 in combined AI credits
Teams breaks even when you need SSO or admin visibility AND you’re a heavy-usage team. If you’re a startup of three people and nobody cares about usage analytics, five Pro subs ($100/mo combined) beat Teams ($200/mo) and give you more total AI budget. If you’re a 15-person agency with a compliance requirement for SSO, Teams pays for itself in reduced admin overhead.
The trap: assuming Teams is cheaper because “pooled billing” sounds cheaper. It’s not. It’s a control layer you buy at $20/seat above your AI budget. Don’t buy it if you don’t need it.
Tokens, Overages, and the Bill That Just Grows
Cursor’s token rates sit at $0.25/M for cache reads, $1.25/M for input, and $6.00/M for output on standard requests. Once you burn through your plan’s monthly credit, on-demand billing kicks in at the same rates. No surprise markups, but no warning either. The bill just keeps climbing.
A single multi-file refactor touching a 50K-token context window can cost $0.15 in input tokens alone. Do that five times in a day and you’ve blown a $20 Pro budget on input costs. Output is where the real spend lives — if Cursor generates 10K tokens of code, that’s $0.06 out of your pocket. Most substantial features cost $0.30 to $0.80 per completion.
The play: Pro users should monitor their usage weekly. If you’re consistently hitting the ceiling, upgrade. If you’re blowing through credits on low-yield tasks, kill Auto mode and switch to Gemini for routine completions. Enterprise users should negotiate a pooled contract so one expensive task doesn’t bankrupt the account.
Our Call
Hobby: Free tier is real. 2,000 tab completions is enough for a few hours of afternoon tinkering, not a workday. Use it as the trial before paying.
Pro ($20/mo): Buy it if you code 10–15 hours a week and don’t mind switching to Gemini for routine work. The moment you hit the ceiling three weeks running, upgrade.
Pro+ ($60/mo): The sweet spot for solo developers on a salary. Enough budget to use Claude for everything without thinking, with room for agents and long context. We’ve spent six months in Cursor and Pro+ stays on until something materially better ships.
Ultra ($200/mo): Only if you’re freelancing and billing hourly, or if you’re running multi-hour agent tasks that chew through 20K+ tokens per task. For personal use, this is leaving money on the table.
Teams ($40/seat): Buy if SSO or admin analytics justify $20/seat above your actual AI budget. Otherwise, issue individual Pro or Pro+ subscriptions and save the admin headache.
Enterprise: Don’t bother reaching out unless you have 30+ developers and a security team. Cursor negotiates pooled budgets and compliance add-ons, but the minimum contract floor is steep.
The final math: Cursor’s pricing is transparent, but it rewards people who read the fine print. The credit-based model is fairer than the old request ceiling — you only pay for what you use — but it demands active model selection. Pick Claude every time and you’ll leave Gemini discounts on the table. Pick Gemini for everything and you’ll miss the performance tier you actually needed.
For a deeper honest take on whether Cursor is worth the money at all, we’ve got the full six-month review. If you’re torn between Cursor and Windsurf, the subscription math shifts in Windsurf’s favor at the Pro+ level. And if agents are the reason you’re looking, Cursor 3’s background agents are genuinely compelling — worth the Pro+ tax on their own.
The bill is real. Do the math before signing up.
What we don't know is documented at the end of this article. We update when we learn more.