⬢github TypeScript · 64 ★ +63 since we first saw it · pushed 7 h ago · MIT
composo-ai/DishwasherBench
A dishwasher-loading game on a physical simulation of one real machine, built to benchmark language models and playable in a browser. The game inside is called Spotless.
DishwasherBench is a browser game called Spotless where you load a physics-simulated Bosch dishwasher and score points for items that come out clean after a simulated wash. It doubles as a benchmark for language models: models play the same game via an MCP server, and episodes are deterministic and replayable so results can be verified.
Why now: It was just released (v0.1.0) and featured on Hacker News as 'Can Frontier Models Stack a Dishwasher?', framing a playful but physical benchmark comparison of LLMs.
Who it is for: LLM researchers and engineers wanting a novel physical-spatial benchmark, plus curious players who enjoy the browser game.
ai-agentsbenchmarkevaluationgamellmllm-benchmarkmcpphysics-simulationsimulationtypescript
Stars over our 145 snapshots: 1 to 64, since 1 d ago.
Where people talked about it
API: https://socialmediatrends-api.osmike.com/v1/repos/composo-ai/DishwasherBench