One brief, every engine, published as it landed
We pay for the renders, run one identical prompt through a slate of engines that spans every price tier, and publish everything: the outputs, the wall-clock times, the credits charged and the engines that failed.
The runs
Evidence, not rankings.
Each page prints the prompt verbatim, so you can reproduce the run on your own account and disagree with us.
AI video model benchmark
One brief, twelve engines from free to flagship. Every clip published, including the wobbly ones, with what it cost and how long it took.
AI image model benchmark
The same product brief through eleven image models, one attempt each, with the price, the time and the pixels you actually get back.
Why bother
Because everyone else cherry-picks.
Model marketing is a showreel: best-of-twenty frames, hand-tuned prompts, no mention of the wait or the price. That is fine as advertising and useless as procurement. If you are choosing an engine to spend money in, you need the median attempt, the failures and the bill.
So we run the same brief through a slate that spans every price tier, once each, and publish the lot. It costs us real credits and it makes some of our own models look bad — including a few that do not work at all, which we leave in the table rather than quietly dropping.
Test them on your own brief
Every model in these runs sits in the same picker on one balance. Free account, 22 credits a day, no card.