I ran local LLMs on Intel's cheapest iGPU, and the results were surprisingly decent
…though it’s the most beginner-friendly option for hosting local LLMs. Its heavy performance overhead already makes it a terrible option for such weak hardware, and it just isn’t flexible…