Local benchmark for tiny System 1 (decision) and System 2 (generative) models: dataset, leaderboard and the models on it.