nomel gave a good answer in a different thread > This is not about the model, it’s about the relative speed improvement from the hardware, with this model as a

nomel gave a good answer in a different thread > This is not about the model, it’s about the relative speed improvement from the hardware, with this model as a demo. To compare apples to apples look at the tokens per second of other systems running Llama 2 70B 4096. We're by far the fastest! https://news.ycombinator.com/item?id=38742466 https://news.ycombinator.com/item?id=38742466