I work on LLMs and it's my setup (I have M1 32gb, though it has little influence as everything is done on colab/server jobs).

Local decice can have more influence if you plan to use the laptop for testing/inference colab made model.