Alibaba Qwen3.8-Flash-Next runs locally on Framework Desktop

The author shares their practical experience running Alibaba's open-weight Qwen3.8-Flash-Next model locally on a Framework Desktop with 128 GB of shared memory. Operating at 4-bit precision, the 125-billion-parameter model runs at roughly 38 tokens per second and handles parallel conversations, though it suffers from poor factual recall unless paired with web search. The author plans to use the local model for heavy-lifting volume tasks while reserving paid models like Claude for final summaries. Open article →