ai tools, ranked
by someone who read the licence

each entry is built from primary sources: the vendor's own pricing page, the actual licence file, and the product documentation, all checked on the date shown at the top of the ranking. where a category has credible independent measurement — public leaderboards with large vote pools, published adoption figures — we use it and say so. these are editorial rankings informed by hands-on use, not the output of a controlled laboratory benchmark, and we don't publish measured figures we didn't produce.

advertising, and affiliate links on some outbound vendor links, where we may earn a commission if you sign up after clicking through. commercial arrangements have no input into ranking, scoring or copy, and no vendor sees a ranking before it publishes. where we operate a tool ourselves, the entry carries a 'we own this' badge and a disclosure on its face rather than buried in a footer.

it weighs capability, operational reality — reliability, latency, support, how bad the failure modes are — and what the tool costs at a realistic volume rather than at its headline entry tier. viability counts too: something with a published shutdown date, a licence you can't ship on, or no api to build against ranks below its raw capability, and the entry says that is why. scores are only comparable within a category, never across them.

often not, and checking is the single most useful habit to copy from this site: open the licence file rather than trusting the launch post. it is common to find downloadable models and tools whose announcements imply permissive terms while the actual licence is non-commercial, caps commercial use above a revenue threshold, restricts competing products, or excludes entire jurisdictions. every ranking lists the licence we found for each entry, and flags where a vendor's own documentation contradicts itself.

ai is the focus — models, agents and the infrastructure around them — because that is where the choices are hardest and the information ages fastest. but we also rank the productivity and operational software that sits alongside them, and we add a category whenever we can research it properly rather than compiling it from marketing pages.

we re-check a ranking when something material happens — a new version, a price change, a licence change, an acquisition or a shutdown announcement — rather than on a schedule. the date at the top of each ranking is when we last went through every figure on that page. if it looks old, treat it as old.

yes, and quickly. prices, limits, licence terms and version numbers move fast enough that some of this will be out of date before you read it. corrections are welcome from anyone, including the vendor being reviewed, and we'll fix a factual error faster than we'll defend it. what we won't do is change a ranking because someone didn't like where they landed.