Multiple AI models are assigned roles and argue a topic to conclusion. A Red Team model then tries to poke holes in the result, and the whole thing is scored across 7 metrics. Currently available as a Beta.

Choose between Debate + Synthesis + Red Team, Tree of Debate, and Issue-by-Issue.
Assign roles like for, against, critic, and judge across 2 to 5 models, mixed and matched from 7 role types.
The quality of the debate is scored automatically across 7 metrics, so you can see how strong the conclusion really is.
A dedicated rebuttal model raises its objections after the conclusion has taken shape — not before.
Set the number of debate rounds anywhere from 1 to 10 (4 by default).
The conclusion, open questions, and constraints can be handed straight to Chat or Compare.
Enter the question or decision you want examined.
Models are assigned for, against, judge, and other roles based on the algorithm you chose.
Follow the argument through the configured number of turns.
A Red Team model challenges the conclusion, and the debate is scored across 7 metrics.
Continue the conclusion as a Chat, or lay it out again in Compare.
Debate mode (Beta) is ready to try as soon as you create an account.
Start free