Test which prompt or model works best

Sets up side-by-side experiments with different models, temperatures, or prompt phrasings, runs them against a test set, and shows you which one wins.

Best for: Builders who suspect a small tweak could improve results but want proof before deploying.