My recommended use on your setup:
Model	Use case
Llama3.1:8b	Main general-purpose assistant
Mistral:7b	Fast, concise replies & RAG
Qwen2.5:3b	Lightweight, quick lookups
Qwen2.5-Coder:7b	Dedicated coding tasks
Llama3:8b	Legacy/benchmark (optional)
qwen2.5:7b-instruct Writing up emails
deepseek-r1 (chonky but accurate)
deepseek-r1:8b (lighter version of r1 , can run on DS1823xs+)