Only the good stuff, picked for you.
The flat monthly options, ranked by whether the ceiling is somewhere you will actually reach. A subscription only beats the meter if you would have spent more than the fee using it.
The default consumer subscription, and the one most people should price everything else against. It buys access rather than tokens, so the maths is about how often you hit the limit rather than what you send.
The tier above, priced for people whose time is worth more than the difference. If you are not routinely hitting the ceiling of the cheaper plan, this is a subscription you are donating to.
The consumer plan that most often gets recommended for long documents, because the window is large and the interface does not fight you about it.
For people who use Claude as their working tool all day and keep running into the cheaper plan. Two tiers, and the difference between them is how much you expect to use it.
The subscription that is really a storage plan with a model attached: the same fee covers the drive space and the assistant, which changes the comparison if you were paying for either already.
The individual tier for the tool most teams already have. Worth checking whether your employer already covers it before paying for it yourself.
Priced per seat with a usage allowance, which is the model most editor agents have landed on. The thing to read before subscribing is what happens when the allowance runs out.
A research subscription rather than a writing one: the value is in the citations and the follow-up questions, not in the prose.
One subscription across several vendors, which is the straightforward reason to pick it: you are buying the breadth rather than the best model of any single one.
Prepaid credits across a very wide model catalogue, which makes it the cheapest way to try something without opening an account with its provider. You pay a margin for that convenience.
Prepaid, pay-as-you-go, no monthly floor. The right shape if your usage is spiky, and the wrong one if you would rather never think about a meter.
The meter everyone else is compared against, partly because the tiering is so wide: the cheapest and the most expensive model on the same price list differ by a factor of five hundred.
The free tier is the most generous of the large providers, which makes it the obvious place to prototype before deciding whether you need to pay anyone.
Prepaid like the rest, but with a peak and off-peak rate, which is unusual enough to be worth planning around: the same batch costs half as much outside the busy window.
The European consumer subscription. The reason to pick it over the American ones is the same reason to pick the API: where the data is processed.
Pay-as-you-go with a free experimentation tier. The pricing is not the argument; the deployment region is.
Bundled with the social platform, which is the honest way to price it: you are buying the integration as much as the model.
A long-context assistant sold on document volume rather than on model size. Cheap enough in China that the free tier is often the whole answer.
Domestic credits with domestic invoicing, which in corporate China is often the deciding factor rather than the price per token.
The hosted version of a tool you could run yourself. You are paying to not operate it, and the community edition is the honest alternative if you would rather.
The ranking is ours. Anything on this page that can be checked is checked against a named source, and the date we last opened that source is printed with it. If a figure and our verdict disagree, trust the figure and follow the link.
A model is listed only once its provider publishes pricing and documents an API. Every figure carries a source and the date we checked it; the ranking is ours.