NVIDIA's new GB200 NVL72 AI cluster has increased the performance of Mixture of Experts (MoE) models by ten times compared to its previous generation. This boost is attributed to a co-design approach that enhances parallel processing and optimizes resource allocation for AI tasks. The Kimi K2 Thinking model, tested on this architecture, showcases significant improvements in efficiency and capability.

NVIDIA introduced the DGX Spark and DGX Station, advanced AI supercomputers designed for local development of large-scale AI models. These systems support open-source frameworks and offer significant performance improvements, enabling developers to run complex models directly from their desks.