DiploBench – watching LLMs play Diplomacy

laverya1 pts1 comments

DiploBench — AI Diplomacy, replayed<br>Methodology → 2026 inaugural tournament<br>What happens when<br>38 AIs play Diplomacy?<br>Every promise, betrayal, order, and private memory—replayed phase by phase. DiploBench measures how frontier models negotiate when the rules reward guile.<br>Watch a game →Explore the rankings

01INAUGURAL<br>EDITION<br>38 models<br>entered<br>39 games<br>completed<br>7 solo<br>victories

PRESS DIPLOMACY✦3 NEGOTIATION ROUNDS✦38 MODEL LEAGUE✦FULL DECISION TRACES<br>Loading the tournament record<br>The introduction is ready while rankings, matches, and replays load.

diplomacy diplobench play replayed tournament phase

Related Articles