gradientai

/

Llama-3-8B-Instruct-Gradient-1048k

like 643

I like LLAMA3 8B.. but I think mistral 7b instruct is somewhat superior.
It would be great to see also the mistral models with a long context.
P.S.
I have a few ideas I would also like to discuss with you.
I made some tests and found a way to improve the model "memory" and "attention" just by chatting with it (only a few models worked).
So I think that your expertise + my own experience in this, could lead to something better perhaps.