Larger reasoning models improve accuracy, but not token efficiency
Scaling DeepSeek-R1-Distill models makes them better at solving harder arithmetic and algorithmic problems, but does not make their correct answers cheaper in tokens. The study models correctness and output length across four problem classes: accuracy falls roughly exponentially as problems grow harder, while answer length follows a power law whose parameters do not systematically improve with model size. Bigger models help, with diminishing returns, but naive scaling leaves efficiency largely unchanged.