I run same local LLM on an RTX 5070 and integrated graphics, and the results were beyond my expectations

You don't need an expensive GPU to start running local AI.