SWE-Game Evaluates Coding Agents on Godot Game Development
October 6, 2026
SWE-Game is a benchmark comprising 247 tasks across 41 executable Godot games. It tests agents on game design implementation, bug repair, and engine porting using engine-state checks and reference-input replay.
HOW THIS AFFECTS YOU
●
builderYou can test if your coding agents are capable of handling complex, stateful software engineering tasks.
●
researcherThe benchmark provides a more rigorous way to evaluate agents using executable end-to-end environments.