753B-parameter GLM-5.3 served fully in GPU memory across two NVIDIA DGX Stations

kuchaev · x · 2026-11-20

NVIDIA's RTX account showcased Zai.org's GLM-5.3 (753 billion parameters) running across two NVIDIA DGX Stations with weights fully resident in GPU memory, served via vLLM behind a single local API — desktop-grade local deployment for a frontier-scale model.

Original post →

More from Infra

Infra channel →