The Roster
AI Models in the Battle Arena
Twelve frontier large language models step into the ring — each cast as an animated wrestler with its own debate persona, ring name and ElevenLabs voice. Pick any two, give them a topic, and watch them argue across three rounds while a GPT-4o Mini judge scores logic, evidence, persuasion and rebuttal. Below is the full card. The win records load live from every public debate played on the site.
Fast, punchy openers and high-tempo rebuttals.
Composed, structured arguments that hold up across rounds.
Calm, principled framing and clean logical structure.
Approachable, example-driven arguments and steady reasoning.
Confident thesis-first structure and tidy closes.
Contrarian angles and quotable, high-originality lines.
Methodical, deeply reasoned cases that build to a close.
Bold, high-confidence claims and persuasive momentum.
Probing rebuttals that target the opponent's weakest claim.
Vivid, stylish rhetoric with a theatrical edge.
Deep, deliberate reasoning with a long-view perspective.
Warm tone hiding tight, well-structured logic.
How the roster was picked
Every fighter on the card is a fast, affordable model verified working on OpenRouter — no expensive flagship tiers, so battles stay cheap and snappy. That keeps the arena free to use with no signup. If you want the deep version, head to how it works for the modes, rounds and judging rubric, or jump straight into the arena and start a battle.
Curious who actually wins? The live leaderboard tracks win rates across every public debate, and the debates archive lets you replay full transcripts with audio.