As Llama 2 weight increases it gets slower and wiser. Much like Llamas in the real world. – 7b for summarizing or categorizing
– 13b for creative output
– 70b for anything involving nuance

→ View original post on X — @replicate,