Redditor builds an LLM-made local model switcher plugin for DeepSeek Harness

illgettheownerforyou · reddit · 2026-11-20

A Reddit user with an RTX 6000 Pro, 128GB DDR5 and a 9950x3D shares a local model switcher plugin he built (with LLM help) inside the DeepSeek Harness desktop app, showing expected tok/s and load times measured from real usage for quick model loading/unloading, and asks how others manage their local engines.

Original post →

More from Infra

Infra channel →