There was no jump: ChatGPT Plus still costs $20/mo, and the $200 figure is a separate tier called ChatGPT Pro, not a price increase on Plus.
If you saw "$200" next to ChatGPT and assumed your Plus subscription had quadrupled in price, you were almost certainly looking at the Pro plan, which OpenAI sells alongside Plus rather than instead of it.
According to our AI tool database, ChatGPT's current lineup is Free (GPT-4o mini), Plus at $20/mo, and Pro at $200/mo — three tiers that coexist. The same database lists Claude as Free, Pro at $17/mo annual or $20/mo monthly, and Max from $100/mo, which shows this is an industry pattern rather than something OpenAI did on its own.
A high-priced top tier sitting above an unchanged mid tier is now normal across the major assistants.
The mechanism behind that structure is worth understanding, because it explains why the $20 tier tends to stay put. Chat plans are sold on a subscription model, where you pay a flat amount and the provider absorbs whatever compute you consume. A small number of users consume enormous amounts — long agentic coding sessions, huge documents, repeated image generation — and on a flat $20 plan those users can cost more to serve than they pay.
Rather than raise the price for everyone, vendors add a high-priced tier that heavy users self-select into. The cheap tier stays cheap because most people stay on it and remain profitable; the expensive tier exists to stop the heaviest users from being subsidised by everyone else.
That is why the number you see at the top of a pricing page often has nothing to do with the number you actually pay.
A concrete example makes this clearer. Say you write marketing copy a few times a week, summarise the occasional PDF, and use the assistant to draft emails. You are a light user.
Plus at $20/mo covers that comfortably, and Pro at $200/mo would buy you nothing you would notice — you would be paying ten times as much for headroom you never touch. Now change the inputs: you run a coding agent for hours at a time, feed it a large codebase, and let it iterate on its own.
That is a genuinely different workload, and it is the kind of usage the top tier is built for. The decision rule that follows is simple and it applies to every assistant, not just ChatGPT. Start on the free tier, use it until you actually hit a limit — a message cap, a context ceiling, a feature that is locked — and only then pay.
When you do pay, count how often you hit the ceiling in a normal week. If it is once or twice, the $20 tier is your answer. If you are hitting it daily, the higher tier may finally earn its price.
Where this advice breaks down is worth being honest about. First, pricing in this category changes constantly, and plan names, limits, and prices get restructured without much warning — the vendor's own pricing page is the only reliable source on the day you buy, and anything written earlier can be stale.
Second, the tiers are not just about volume. Some capabilities are gated to higher plans regardless of how lightly you use them, so a light user who needs one specific feature may still have to pay up, which breaks the "count your limit hits" rule. Third, annual billing can change the maths in ways a monthly comparison hides — Claude's Pro tier, for instance, is $17/mo on annual billing versus $20/mo monthly, a real discount for committing upfront, but it also locks you in if your needs change.
Fourth, none of this tells you whether the assistant is any good for your particular job; a cheaper plan that fits your workflow beats an expensive one that does not. If you are weighing a $20 chat subscription against other tools in your budget, it helps to see how the whole stack adds up rather than judging one line item in isolation.