GLM5.2 on AMD MI355X at 2626 tok/s/node at over 2x lower cost than Blackwell

In a recent post, the author 'latchkey' discusses the performance of the GLM5.2 model running on AMD MI355X hardware. The model achieves an impressive speed of 2626 tokens per second per node, while operating at a cost that is more than twice lower than that of Blackwell. This advancement could significantly impact the efficiency and affordability of AI model deployment in various applications. For more details, visit the original post here.