Local LLM speed perception

Verify the prefill and token generation speed for the selected hardware and model.