My hope is that they are taking longer because of a memory system upgrade that will make running significantly more powerful LLMs locally more feasible.