Tips for optimizing local LLM execution with Ollama (Qwen 3 8B) on low-RAM machines

#Good afternoon,

I have recently started running workflows in KNIME that connect to a local LLM (Qwen 3 8B) using the Ollama integration on Windows.

I am aware that my setup is quite limited—my computer only has 8 GB of RAM—and execution is currently rather slow.

Could anyone share some recommendations, best practices, or lessons learned on how to optimize local LLM execution in KNIME or run similar models efficiently with limited hardware resources?

Thank you very much in advance!