I tested 8 LLMs as tabletop GMs – a 27B model beat the 405B on narrative quality
Sum B+a+c+k+g+r+o+u+n+d: I've been working on an open source agentic tabletop GM as a leisure project intended to run on any LLM with tool support. I started it as a Claude Code skill to run D&D sessions and eventually generalized it to b…