Qwen 3.6 27B excels as local development model

Qwen 3.6 comes in MoE (35B A3B) and dense (27B) variants, with the 27B model recommended for local use. It runs on llama.cpp and performs well on coding and creative tasks, often outpacing expectations for its size.