Mode (v2.1.36+) running 2.5x faster at 6x the token cost, on the same underlying model. Toggle with `/fast`.