← All reviewed products
Edition 1 · Updated · 51 products

Cursor pricing, reviewed

How Cursor has priced its AI from January 2025 to now — the rung, the moves, and where it's headed.

  • CompanyAnysphere
  • CategoryCoding agents
  • AudienceProsumer
  • PatternCredit burn-rate reprice
  • Last observedJun 2026
  • Confidencehigh
  1. Jan 2025 Rung 2 · Soft meter
  2. Jun 2025 → now Rung 4 · Metered overage Pro reworked to compute/$ limits; pay-as-you-go overage becomes the default

Moved across the continuum this window — pattern: Credit burn-rate reprice.

The short version The sticker barely moved — but the meter did. The same work costs more than it used to.

The verdict

In January 2025, Cursor's $20 Pro plan was effectively a soft meter — a few hundred fast requests plus unlimited slow ones, so running out only dropped you into a throttled queue, not a paywall. On June 16, 2025 Anysphere reworked Pro around compute-and-dollar limits, stripped the unlimited fallback from the frontier models, and made pay-as-you-go overage the default above the allotment — all while the headline price stayed $20 and effective coverage roughly halved. The fierce backlash forced a July apology and refunds, but the usage-based model was kept, not rolled back. Cursor sits today at metered overage.

Current pricing snapshot

As published on Cursor's own pages · captured Jun 12, 2026. Unit: USD of included model usage (metered at API rates).

PlanPrice / moIncludedAt zeroTop-up vs plan
Hobby Free Unstated
Pro $20 20 Opt-in overage Yes · price hidden
Pro+ $60 70 Opt-in overage Yes · price hidden
Ultra $200 400 Opt-in overage Yes · price hidden
Teams Standard $40 Opt-in overage Yes · price hidden
Teams Premium $120 Opt-in overage Yes · price hidden
Credit ↔ dollar
Pegged — one credit is a fixed $ amount
Credit burn rate
Varies by model / action
Model choice
You pick, or auto-routed
Billing
Monthly or annual
Purchased credits
Unstated

What changed

  • Jun 16, 2025 · Credit burn-rate repricing · Official: Pro moves from request limits to compute/dollar limits — $20 of model inference at API prices/month; Ultra ($200/mo, 20x usage) launched; 'Auto' model made unlimited but everything else metered. Same $20 price, but effective coverage roughly halved and the unlimited-slow safety net was removed for frontier models — overage now auto-bills at API rates in arrears.

Where it's headed

Already on the overage rung, the gravity points toward pricing the action directly and thinning the seat floor. Watch for further granularity changes to the meter rather than any return to flat pricing.

A read on the likely next move from where Cursor sits today — association, not a forecast (the panel is observed, not sampled). How the continuum works →

Sourced from Cursor's official pricing pages and archived snapshots. Primary: pricing page. Method & limits: Methodology & data.