Making a non-scientific benchmark with Python and Llama-CPP
![Towards Data Science](https://miro.medium.com/v2/resize:fill:48:48/1*CJe3891yB1A1mzMdqemkdg.jpeg)
About a week before the time of writing this story, new open Llama-3 models were released by Meta. As claimed by Meta, these are “the best models existing today at the 8B and 70B parameter scales.” For example, according to a HuggingFace model page, Llama-3 8B got a…