/grok how many tokens per second does a CPU-only server process and generate(Xeon E5 processors) vs GPU servers? How long will it take? Will it be enough for a forum assistant like you? How efficiently will it read context? ( I have read that it processes 100-200 tokens per second and generates...