For model makers

Embeddable badges

A badge shows where a model stands in our independent tests: its rank in a use case, its rank overall, or a model picker award. Every badge links to the page with the evidence.

Badges come from the same rankings and the same model picker as the rest of the site: every model gets the same prompts with the same settings, scored blind. How we test

Badges update themselves. When a new test changes a result, the embedded badge shows the new standing and the month of the data.

#1 by use case

Picker awards

From the model picker: the cheapest model near the top score, the fastest strong model and the most accurate text, per use case.