Minimum VRAM requirement for 1M context length and recommended hardware?
#6
by Kinfai - opened
Hello,
I am planning to deploy the GLM-5.3-NVFP4 model and want to utilize the full 1 million token context window. Could you please provide some guidance on the hardware requirements?
Specifically, I would like to know:
What is the absolute minimum VRAM required to run this model with a 1M context?
Do you have a recommended typical hardware configuration (e.g., number and type of GPUs) for optimal performance at this context size?
Thank you in advance for your help!