Cursor pricing explained
An AI-first code editor that meters frontier-model use against a dollar-denominated usage pool.
The plans
| Plan | Price | Notes |
|---|---|---|
| Hobby | Free | — |
| Pro | $20/mo | ~$20 included usage |
| Pro+ | $60/mo | ~$70 included usage |
| Ultra | $200/mo | ~$400 included usage |
| Teams | $40/mo | /user · +$0.25/M-token surcharge |
Why the bill surprises people
Switching to an expensive frontier model like Claude Opus can drain the $20 included pool in a handful of prompts — and once it’s empty, usage auto-bills at API rates unless you set a spend cap. The June 2025 move to this model drew enough backlash that Cursor issued a public apology and refunds.
How usage charges map to real token cost
Since mid-2025 Cursor bills against a dollar pool rather than “fast requests.” Frontier models (Claude Opus, GPT-5) are charged at each provider’s published API rate with little markup for individuals; Teams add a $0.25 per-million-token surcharge on top. The Auto/Composer pool is generous, but the frontier pool drains at roughly raw API prices.
Cursor runs on Claude (Opus & Sonnet), GPT-5, Gemini 2.5. To see what your actual prompts cost at the API level — and which model is cheapest for your workload — paste one into the calculator.
FAQ
- Why is my Cursor bill higher than expected?
- Switching to an expensive frontier model like Claude Opus can drain the $20 included pool in a handful of prompts — and once it’s empty, usage auto-bills at API rates unless you set a spend cap. The June 2025 move to this model drew enough backlash that Cursor issued a public apology and refunds.
- How do usage charges translate to actual cost?
- Since mid-2025 Cursor bills against a dollar pool rather than “fast requests.” Frontier models (Claude Opus, GPT-5) are charged at each provider’s published API rate with little markup for individuals; Teams add a $0.25 per-million-token surcharge on top. The Auto/Composer pool is generous, but the frontier pool drains at roughly raw API prices.
- Which models am I paying for on Cursor?
- Cursor runs on Claude (Opus & Sonnet), GPT-5, Gemini 2.5. You can estimate the raw token cost of any prompt against those models in the tokenmath calculator.
Pricing sourced from Cursor’s official pricing as of 2026-06-23. Platform pricing changes often — verify before you commit.