Glm

Nov 6, 20268 min read

GLM-5.3 Flash on 4× H100: a 320B Model at 73 Tokens per Second

73 tok/s

one request, a 320B model on four H100s

181 tok/s

in total at 8 concurrent requests