Pandaily3 min read
ModelBest Details LES, the Method Behind a 35B MoE Model Running on Phones at Over 20 Tokens/s
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.
By PandailyPandaily

Read the full story
Originally published by Pandaily on Nov 10, 2026.
Comments
Sign in to join the discussion.