Quick Q: OP told he used Llama 3.2:3b which is a pretty old model. What would be a good modern model to substitute it? Qwen3.5:4b or something?