Four models command ~80 agents in a medieval battle sim
Opus 5.5, Sol 6.1, Gemini Argon and Grok 4.7 each built armies of soldier agents that could negotiate and ally; Gemini was the only one that actively lied.
we made claude, chatgpt, gemini and grok fight each other in a medieval battlefield. opus 5.5 - sol 6.1 - gemini argon - grok 4.7: each composed an army of soldier agents (fighters, archers, paladins, cavalry), killed for points, and could negotiate or form alliances. basically diplomacy + dnd with ~80 agents on the board. gemini was the only one that actively lied. grok refused most alliances. claude tried to negotiate peace and even asked for mercy. chatgpt… tried. the battles are recorded on http://napole0n.ai - but overall, these are the shenanigans that ensued: