The Lenovo ThinkSystem WA7785a G3 server has achieved a record-breaking throughput of 6708 tokens per second while running the 671B parameter DeepSeek large language model. This demonstrates the exceptional performance capabilities of the system for demanding AI workloads.