AINo affiliate program

Cursor pricing, and the tier that pays you back

Cursor costs $0 on Hobby, $20 on Pro, $60 on Pro Plus and $200 on Ultra per month, plus $40 or $120 per user on Teams. The plan price is the wrong number to compare: what separates the tiers is how much third-party model usage each one includes — and only one of them returns more than you hand over.

Category: AI
Affiliate: None. Independent review
Coverage: 31 articles

The plans · updated 31 July 2026

Four tiers. One that pays for itself.

Individual pricing runs from free to $200 a month. What the tier list does not show is the model usage each one includes, which is the number that decides the bill.

Hobby

Free

Try the agent without a card.

Limited agent requestsAccess to ComposerNo credit card required

Where it runs out

Pro

$20/ mo

Includes $20 of third-party model usage — exactly what you pay.

$20 of Other Models usageGenerous Cursor Models usageCloud agents, MCP, skills, hooksUnlimited tab completions

See the ratio

Pro Plus

$60/ mo

Includes $70 of model usage. The first tier that returns more than it costs.

$70 of Other Models usage3× Pro agent limitsCovers the $60–$100 daily-agent bandCheaper than Pro plus overage

Why this one

Ultra

$200/ mo

Includes $400 of model usage — double the subscription.

$400 of Other Models usage20× Pro agent limitsPriority access to new featuresFor multi-agent workloads

See the ratio

Also available

Two plans that sit outside the individual ladder.

01TeamsStandard, or Premium at 5× agent limits. Adds SSO, pooled admin, Bugbot reviews and usage analytics — and a per-token surcharge on third-party models.$40–$120 / user / mo

02Start — India onlyTax inclusive, billed in INR by UPI or card. Cursor's own models only: no third-party pool, no Bugbot, no Auto, no SDK.₹649 / mo

The number nobody publishes

Divide the price by what it buys.

Cursor states what every plan costs and, separately, how much third-party model usage each includes. It never divides one by the other, and neither does any page currently ranking for this query. Do it and the ladder stops being a ladder: Pro returns exactly what you pay, Pro Plus a little more, Ultra double.

1.00×

Pro returns what you pay

$20 in, $20 of model usage

1.17×

Pro Plus returns slightly more

$60 in, $70 of model usage

2.00×

Ultra returns double

$200 in, $400 of model usage

$0.25

Per-million surcharge on Teams

Third-party models only

Why the bill surprises you

Two pools, not one balance.

Cursor bills against two separate allowances that both reset on your billing date. Conflating them is the most common reason a plan feels smaller than advertised — the plan has not shrunk, you simply picked a model with a higher per-token rate.

What draws from what

Cursor Models

Grok 4.5 and Composer 2.5, Cursor's own. Generous included usage on every paid tier, and exempt from the Teams surcharge.

Other Models

Claude, GPT and Gemini, charged at each model's real API rate against the dollar figure your plan includes.

On-demand

Past the included amount you continue at the same rates, billed in arrears. Requests are never slowed or downgraded.

Start (India)

Cursor Models only. There is no Other Models pool on this plan at all, and no on-demand usage.

What draws it down

The rates behind the allowance.

The spread is wide enough to change which tier you need. At Claude Fable 5's output rate, Pro's included $20 buys roughly 400,000 output tokens. At GPT-5.6 Luna's, the same $20 buys over sixteen million.

Model	Input	Cache read	Output
Grok 4.5	$2	$0.50	$6
Composer 2.5	$0.50	$0.20	$2.50
Claude Sonnet 5	$3	$0.30	$15
Claude Opus 5	$5	$0.50	$25
Claude Fable 5	$10	$1	$50
Gemini 3.1 Pro	$2	$0.20	$12
GPT-5.6 Sol	$5	$0.50	$30
GPT-5.6 Luna	$0.20	$0.02	$1.20
Auto Cost — any model	$1.25	$0.25	$6

Per million tokens, from Cursor's published model table. Cache-write omitted where unpriced.

The decision

Match the tier to the usage.

Cursor publishes its own usage bands, and they line up with the ratio in a way worth acting on. The middle band is the one that costs people money: a daily agent user is told to expect $60–$100 of usage a month — more than Pro's $20 allowance, and inside Pro Plus's $70.

By how you work

Cursor's own usage bands, matched against what each tier includes.

01Daily tab completionsStays inside $20. Hobby may genuinely be enough.Hobby / Pro

02Occasional agent runsUsually lands inside the included $20.Pro

03Agents every day$60–$100/mo of usage. Paying overage on Pro costs more than the upgrade.Pro Plus

04Multiple agents, automations$200+/mo, and the only tier where the ratio doubles.Ultra

Not on the pricing page

The Teams token surcharge.

On Teams and Enterprise, third-party model requests carry a Cursor Token Rate of $0.25 per million tokens on top of the model's own API price. It applies to included usage, to on-demand usage, and even to your own BYOK keys.

Cursor's own models are exempt, and so is Auto Cost. At team volume that makes routing work to Grok 4.5 or Composer 2.5 a real saving — and none of it appears on the pricing page.

Questions