StarSkirmish Bench

StarSkirmish Bench is a platform that evaluates large language models (LLMs) by having them write Protoss bots for StarCraft: Brood War in C++. Each LLM gets one hour to write a bot which is then tested against a range of human and AI bots. The benchmark measures long-term strategic coding and reasoning abilities of LLMs through gameplay performance scaled from weak demo bots (0) to top human bots (100). It provides detailed ranking by Elo ratings and tracks model effectiveness and efficiency.