552B DeepSeek V4.1-Flash Model Achieves 494 Tokens/Second Output

The 552B DeepSeek V4.1-Flash Model achieves a peak output of 494 tokens per second when powered by an at-home rig consisting of four NVIDIA DGX Spark units. This performance metric highlights the model's efficiency and capability in processing large amounts of data quickly. The use of NVIDIA DGX Spark units, known for their high computational power, allows users to leverage advanced AI capabilities in a home setup. This development may influence AI engineers and researchers looking to optimize their hardware configurations for enhanced performance, particularly in natural language processing tasks. This follows DeepSeek's recent advancements in narrowing the AI technology gap with the U.S., as reported earlier this month.