Comment by tintor

Airgapped LLM inferrence server can't serve their output tokens, right?