glm 1

• 2x GH200 for LLM inference, Part 3: GLM-5.2, expert offload, and the CPU question Jun 17, 2026