Isn't it a challenge today to run a large LLM on a CPU/GPU as those found in mobile phones? I would have thought that only the information that it might be pos

Isn't it a challenge today to run a large LLM on a CPU/GPU as those found in mobile phones? I would have thought that only the information that it might be possible, is a good news?