For model makers
Launching a model? Get an independent test on launch day, run with the same method as everything else on this site.
Launch Lab
Give us early access and we run your model through the jobs it is built for, against the current leaders, with our standard briefs and blind grading. The result is published on launch day, with outputs, scores and cost per usable result.
Ongoing coverage and custom reports
We re-test when you ship new versions, and we can run private test rounds on the jobs your customers care about, in English and Chinese.
The rules
- You never see scores before they are published, and you can't change them.
- Paid work is labelled as such on the page.
- Results are published whether your model wins or loses.
Contact us to talk about a launch.