ParentFull threadsebastianconcpt·Better than Qwen3.6-35B-A3B-8bit ?When I tried glm found it way way slower (omlx as runtime)View on HN