The subscription showdown: Claude, ChatGPT, Cursor and Google.
Four ladders, the same three rungs: a $20 sampler, a ∼$100 working tier, and a $200 ceiling. Which rung — and whose ladder — depends on the work.
The market has quietly standardised. Every provider now sells the same shape: a free tier to hook you, a $20 tier that’s superb for chat and rationed for agents, and one or two power tiers for people whose output pays for the seat. Prices below were checked on 6 July 2026 — they change often, so treat the shape as durable and the digits as perishable.
What each ladder is actually selling
Checked 2026-07-06
If your subscription stack has grown by accretion rather than decision, that’s exactly the kind of line item our Reality Check prices up — including which of these four ladders your team should actually be standing on.
Code all day → Claude Max 5× or Cursor Pro+. Terminal-first, one codebase, deep reasoning: Claude Code. Living in an editor, hopping between models: Cursor. Both working tiers pay for themselves in hours saved within the first week.
Mixed work, one seat → ChatGPT Plus. Still the broadest $20 in software. Just know the demo-reel models live at $200 — and there’s no middle rung when you outgrow Plus.
Live in Gmail and Docs → Google AI Pro. The 1M-token context and Workspace integration make it the best $20 for document-heavy teams — and the quiet May price cuts made its upper rungs the cheapest ceiling of the four.
Don’t stack all four. The standard set costs ≈$110 a month per person and we’ve never audited a business where four parallel subscriptions beat one well-chosen tier plus a free-tier backup. Pick one primary, upgrade on evidence.
Match the tier to the loaded hour, not the sticker. A $100 upgrade that saves a senior person three hours a month is free. Forty $20 seats nobody opens are not. Read the usage dashboards before the renewal, not after.