However the LLM doing inference on 2 H100s can easily exceed 10x the content generation rate of a single human. Regarding quality of the output, it obviously d

However the LLM doing inference on 2 H100s can easily exceed 10x the content generation rate of a single human. Regarding quality of the output, it obviously depends on the task, but we are benchmarking the models on real world metrics and they are beating most humans already.