ModelBest的LES从手机闪存中预取下一层专家;它在iPhone上报告20.51代币/秒,是其基准的6.07倍,并开源了代码。ModelBest's LES prefetches next-layer experts from phone flash; it reports 20.51 tokens/s on an iPhone, 6.07x its baseline, and has open-sourced the code.