Lenovo has proudly announced that its pioneering AMD AI large model training server, the Lenovo Wentian WA7785a G3, has shattered performance barriers by achieving a throughput peak of 6708 tokens per second while deploying the full-scale 671B DeepSeek large model on a standalone machine. This groundbreaking achievement sets a new benchmark for running large-scale models on a single server, underscoring Lenovo's profound technical expertise and swift research and development prowess in the realm of AI infrastructure.
