Local LLM offloading

[ciroiriarte]

Hello,

I'm wondering if with the current state of local LLM tools/models, would it be possible to keep using Opus + oh-my-claudecode for the most demanding design sessions and offload most of the mechanical tasks to a local model.

is that something you currently do?, how can omc fit that model?.

I'm thinking about downgrading to Max 5x and going single or dual RTX 3090.