Composer, Grok, and list-price models

Cursor Cost Calculator

Estimate Cursor token cost for Composer and Grok at the published overage rates, or for Claude and GPT at API list price. Plan stickers are shown next to the estimate, with the limits of that comparison written on the card.

Claude Code Codex Cursor

Model

Cursor's own model. Rates are what Cursor publishes for usage beyond the weekly pool. No cache-write price.

Showing an example: One Cursor agent session on Composer 2.5. Edit a field, or clear it, to price your own usage.

Uncached input tokens

$0.50 / MTok

Output tokens

$2.50 / MTok

Cache reads

$0.20 / MTok

Cache writes

Not published

Repeat this workload

Scales the dollar total after long-context pricing is chosen, so 20 short requests stay on the short-context rate.

AdjustmentsFast modeComposer 2.5 Fast is a separate rate, not 2× standard.Add Cursor Token Rate ($0.25 / MTok)Teams and Enterprise third-party requests only. Cursor models are exempt. Individual plans do not pay this.

API-equivalent cost

$0.75

Checked November 2, 2026StandardExample

Composer 2.5. List rates only. This is not an invoice.

Cost by token type
Type	Tokens	Cost
Uncached input$0.50 / MTok	400,000	$0.20
Cache reads$0.20 / MTok	1,500,000	$0.30
Cache writes$0.50 / MTok	0	$0.00
Long-TTL cache writes$0.50 / MTok	0	$0.00
Output$2.50 / MTok	100,000	$0.25

Against plan stickers

Cursor Plus
$8.00/mo
$0.75 is 9.4% of the $8.00 Cursor Plus sticker
9.4% of the monthly sticker
Third-party models are not on Plus. The weekly pool for Composer and Grok is shared with Grok and is not a token quota.
Cursor Premium
$30.00/mo
$0.75 is 2.5% of the $30.00 Cursor Premium sticker
2.5% of the monthly sticker
Third-party tokens are prepaid credit at API list price, from the first request. They are not included in the $30.
Cursor Super
$100.00/mo
$0.75 is 0.75% of the $100.00 Cursor Super sticker
0.75% of the monthly sticker
Third-party models are still extra, at API list price.
Cursor Ultra
$200.00/mo
$0.75 is 0.38% of the $200.00 Cursor Ultra sticker
0.38% of the monthly sticker
Third-party models are still extra, at API list price.

Assumptions behind the comparison

Subscription plans do not publish a token quota. Session and weekly limits depend on model, prompt length, files, and tools.
The percentage next to a plan is API-equivalent cost divided by the monthly sticker. It is not remaining included usage.
Max and Ultra include monthly API credits that Anthropic says cannot be used in Claude Code or the Claude apps. Those credits are not subtracted.
Claude Pro annual billing is $200 up front (about $17 a month). Comparisons use the $20 monthly sticker.
ChatGPT Pro is monthly only, at $100, $200, or $500. Codex shares usage with the rest of the plan.
Cursor Plus through Ultra include a weekly pool for Cursor models (Composer and Grok), shared with Grok apps. The pool is not a dollar credit equal to the subscription price.
On Cursor Premium and above, third-party models are billed from the first request at API list price, paid from prepaid credit. That spend is not included in the subscription sticker.
The Cursor Token Rate is $0.25 per million tokens on Teams and Enterprise third-party requests. Individual plans do not pay it, and Cursor's own models are exempt. It is off unless you turn it on.
Haiku 5.5's higher rates apply per request whose prompt exceeds 100,000 tokens, including cache reads and cache writes. A summed month of short requests should stay on the standard card. Use Standard rates, or scale with Repeat, so the total is not treated as one giant prompt.
OpenAI long-context rates apply to the whole request when input-side tokens exceed 272,000. Grok 4.7 long context (above 256,000) is 2×, and fast long context is 3×, per Cursor's card.
When a log has a flat cache-write total and no 5-minute / 1-hour split, writes are priced at the cheaper 5-minute rate. Claude Code often uses the 1-hour cache, which costs more.
On Codex logs, cached input is treated as a subset of input tokens. On Claude logs, input tokens exclude cache reads and cache writes.

Prices are approximate public list rates. Anthropic, OpenAI, and Cursor change them without notice. This tool does not call an API. It leaves out tax, negotiated discounts, prepaid credits, batch and flex tiers, and per-call tool fees. A plan comparison is sticker price versus API-equivalent token cost. It is not a promise that the workload fits inside the plan.

Cursor cost for Composer and Grok

Plus is $8 a month and includes Cursor models only: Composer 2.5 and the listed Grok models, from a weekly pool shared with Grok apps. Premium is $30 and about 4× that pool. Super is $100 and about 15×. Ultra is $200 and about 50×. The pool is not published as a token count, so a dollar figure cannot tell you how much of the week is left.

One Cursor agent session on Composer 2.5Composer 2.5 · Standard · total $0.75

Token type	Tokens	Rate	Cost
Uncached input	400,000	$0.50 / MTok	$0.20
Cache reads	1,500,000	$0.20 / MTok	$0.30
Output	100,000	$2.50 / MTok	$0.25

Cursor plan price versus API list price

Third-party models are not in the weekly pool. Premium and above bill them from the first request at API list price, paid from prepaid credit. The $30, $100, or $200 sticker is a separate fee. Plus cannot select those models at all.

The same Claude and OpenAI rates used here are listed on the pricing table. A Claude Code log is usually easier to price on the Claude Code cost calculator.

Cursor token rate and long-context cost

Teams and Enterprise can also pay a Cursor Token Rate of $0.25 per million tokens on eligible third-party requests. Individual plans do not. The toggle adds it only when you turn it on, and never on Composer or Grok.

Grok 4.7 above 256,000 input tokens is 2×, and fast mode above that line is 3×. Grok 4.5 Fast output is $18 per million tokens on the published card, not $12. Cursor does not publish a separate cache-write price for these models. Tokens you enter as cache writes are priced as input, and the result says so.

Related: the Codex cost calculator and the coding agent cost comparison.