how can deepseek be so cheap* yet so effective?

*pricing: MODEL deepseek-chat deepseek-reasoner 1M INPUT TOKENS (CACHE HIT) $0.07 1M INPUT TOKENS (CACHE MISS) $0.56 1M OUTPUT TOKENS $1.68

I think it's because of a combination between the MoE model architecture and the inference done in large batches and run in parallel