Is Cursor's $8 Plus plan (Grok models only) worth it?

Cursor's cheapest paid plan is now Plus at $8 a month ($80 a year). That undercuts GitHub Copilot Pro ($10) and Claude Pro ($20), but there's a catch: you only get Grok models, at settings you can't change. Here's what the $8 actually buys, and who should go straight to Premium.

What Plus includes (sourced)

• Price: $8/month, or $80/year.
• Models: Grok models only. Agent runs Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5 at fixed settings: Fast off, medium effort, Grok 4.7 at 256k context.
• Usage: a base usage pool that refreshes weekly. The pool is shared across Cursor and the Grok apps (Chat, Imagine, Voice, Bot).
• Billing: Cursor subscriptions are now sold through the X Subscription (X Subscription Corp). Plans moved there on Oct 13, 2026, according to PlanDiff. Existing Pro, Pro+ and Ultra customers keep their legacy plans.

What Plus does not include

Per Cursor's docs, Plus has no:

• Auto model selection
• Cloud Agents or Automations
• Bugbot (AI code review) or Origin code hosting
• Third-party models (Claude, GPT, Gemini and so on)
• Bring-your-own API keys

All of those start at Premium, $30/month, which also gives 4× Plus usage. On Premium, third-party models are billed at API list price from prepaid credit.

The ladder, by price per unit of usage

Cursor sells every tier as a multiple of Plus usage. Dividing the monthly price by that multiple gives the cost of one "Plus-sized" pool. This is our arithmetic from Cursor's published numbers, not a test result.

Paying yearly saves $16 on Plus ($80 instead of 12 × $8 = $96). Usage gets slightly cheaper per unit as you go up, but the big jumps are the features Premium unlocks, not the price per unit.

Per-token rates for the Plus models

Cursor's docs list these rates per 1M tokens:

Composer 2.5's output is less than half the price of Grok 4.7's, so which model the Agent picks matters for how long a weekly pool lasts. How long it actually lasts: [TEST RESULT PENDING].

Hands-on: the same build on Plus

Test brief: [BRIEF v1 — browser game: title screen, score, restart]

• How much of the weekly pool one build uses: [TEST RESULT PENDING]
• How many builds fit in one week's pool: [TEST RESULT PENDING]
• Did the fixed settings (medium effort, Fast off) slow us down? [TEST RESULT PENDING]
• Same build on Premium with a third-party model, for comparison: [TEST RESULT PENDING]

Who Plus is probably for (pricing only, not yet tested)

• Leaning yes: hobbyists and vibe coders who already pay for X or Grok, since the pool is shared, and who don't care which model writes the code.
• Leaning no: anyone who wants Claude or GPT models, cloud agents, Bugbot reviews, or their own API keys. None of those are on Plus.

These are early reads from the plan rules only. The verdict below waits for the hands-on test.

Weighing Cursor against the alternatives? See Cursor vs Claude Code vs GitHub Copilot for how the three ration usage, or our GitHub Copilot Pro review for the $10 option.