Exploring Llamafile: Mozilla's Attempt in the World of Open Source AI

https://itsfoss.com/llamafile/

11 points · 4 comments · view on lemmy.world

4 Comments

muntedcrocodile@lemm.ee · 2 pts · 1y (3 replies)

The issue isnt if its easy to use the issue is a matter of compute. A majority of people on the internet only have mobile access. We need a way to let the masses utilise distributed compute in a secure and private way.

JackGreenEarth@lemm.ee · 1 pts · 1y (2 replies)

You can run many 7B models on phones with 8GB RAM

muntedcrocodile@lemm.ee · 3 pts · 1y (1 reply)

I find the 7b models dont have the capabilities id like and i cant image the tokens per second would be very good.

minnix@lemux.minnix.dev · 2 pts · 1y

Yeah running it and it actually being useful are two very different things.