One thing worth trying with General Local AI knowledge guide: improving setup choices is a controlled comparison around general and local. Use the same input for each test and note what changes in the output rather than judging from a single run. If you share the exact setup, it would be easier...