I'm using Mistral 7B 8-bit and it's really fast, my PC is average. The quality might not exceed ChatGPT3.5, but it's pretty decent if you ask me. For most people it's enough, and it's free and offline.

I agree and unless you have a beast computer, the local LLMs are so slow

doofus

I've dabbled with open source models, but have yet to see one that exceeds the quality of ChatGPT3.5. But I get that things are changing fast. Am I wrong?

MattAhlborg0

Mistral 7B 8-bit is pretty decent for most promps. Try it out and let me know what you think. Meta is also open-sourcing their LLama models.

JuanMiguel

You can't self-host GPT-4.  You can't do pay-for-what-you-use without KYC.

elvismercury

PlebeiusG

bitcoin

Introducing PayPerQ (ppq.ai), your new default GPT4 experience, powered by LN

This is pointless. You can already self-host your own open source LLMs on your own machine using LM Studio.