$200 Three-in-a-Row: Cursor Ultra · Claude Max 20x · OpenAI Pro

Not the same kind of "strength" three-in-a-row ~$200 monthly top-tier, everyone calls it "the strongest," but the billing isn't based on the same currency: some are like API pre-paid credits, some are like subscription multiples + cycle caps, some are like quota limits within a 5-hour window. I use all three, just want to clarify the comparison standards:
Cursor Ultra ($200)
The most "API ledger-like" is the $400/month API pool, where selecting a model deducts based on $/M.
Aligning with the same API pricing as per documentation (Claude 4.7 Opus ≈ $5/M in, $25/M out; GPT-5.5 ≈ $5/M in, $30/M out): full input burn ≈ 80 million tokens
Full output burn: Opus ≈ 16 million vs 5.5 ≈ 13 million
(There’s also Auto + Composer in a separate pool, with internal pricing of $1.25/$6/$0.25, official only says it's generous)
Claude Max 20x ($200)
The core is relative to Pro multiples + weekly limits (plus official capacity management space), not fixed token pre-paid packages.
Cannot do a 1:1 conversion with the $400 API pool; just a rough equivalence for reference.
ChatGPT Pro ($200)
Chat side GPT-5 official unlimited (within ToU);
Codex is 300–1600 local messages every 5 hours (local/cloud shared within 5h window), unit is "messages," not tokens, depending on how heavy each message is.
In summary:
All three can be purchased and are all useful. The difference lies in which workflow your main battlefield falls into —
IDE + Agent + selecting model deducts API from Cursor’s "USD → $/M";
Claude full suite collaboration, consuming Max’s cycle rules;
Terminal/Codex heavy users, consuming 5h window + message count limits.
Viewing the three cards separately is more cost-effective and less hassle than asking "who is the strongest."
See the chart below for pricing reference:
· · ·