Methodology

Every plan record is built by hand from primary sources — the vendor's own pricing page, its documentation, and its changelog where a limit is explained. Each record stores the exact URLs it was built from and the date it was last checked, and those links appear on the plan page so you can verify any number yourself.

Usage allowances are quoted in the vendor's own terms. Vendors describe limits as multipliers of a lower tier, credit counts, request counts, token counts, or fair-use language, and we do not convert them into a single invented number — that conversion would be misleading.

A winner is marked only when the row is numeric, the direction is unambiguous (cheaper is better, more bundled credits is better), every compared plan publishes a value, and the winner is unique. Ties, missing values, and rows that are not comparable across vendors — like usage multiples defined against different baselines — stay unmarked.

The finder on the plans page filters the same verified dataset as everything else here. Budget, trial and team filters read published prices and flags directly. Tool and model filters match the vendor's own support lists — a plan is tagged for Claude Code, JetBrains or Claude only when the vendor lists it, and bring-your-own-key wording never counts as a model the plan includes.

Choices inside a row are alternatives: pick VS Code and JetBrains and plans supporting either stay. The switches — trial, team features, open to any agent — must all hold, so ticking one can only narrow the list. Nothing is scored, ranked or promoted, and no vendor can pay to appear under a filter. When a combination matches nothing, the finder says which single change would bring plans back instead of quietly relaxing a constraint for you.

Which matchups get their own page — and the order of the most-searched list on the home page — comes from observed search demand: the autocomplete ladders Google, Bing and DuckDuckGo return for “X vs” and “X vs Y” queries, plus the number of community threads and videos a matchup has attracted. Nothing about a pair's position is paid for.

The comparison tools are the other half of that loop. The pairs people actually build there are the input for which combinations get promoted to their own page next, so the list follows real usage rather than our guesswork.

No vendor pays for placement, ranking, or a winner mark. We do not invent estimates for unpublished limits; missing data is shown as a dash. Arbitrary user-picked combinations live under the comparison tools and are not turned into indexable pages, and vendor names and logos are used for identification only — they belong to their owners, and this site is not affiliated with any vendor.

Plans change often, and prices move. We re-check every record on a regular schedule and whenever a vendor announces a change; the verification date on each page reflects the last check. If something looks wrong, email [email protected] and we will re-verify against the official source.