• mixture-of-experts — A model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.

When HBM binds before compute does, offloading activations to host memory can beat recomputing them by up to 57% — and unlock batch sizes that were previously impossible — provided the transfers actually overlap.

• articleHow NVIDIA Groq 3 LPX Unlocks Ultrafast Interactivity at Long Context on NVIDIA Vera Rubin
• articleNVIDIA AVO Reaches 100% on ARC-AGI-3, Demonstrating a Frontier-Level General-Purpose Architecture for Long-Horizon Autonomous Agents
• articleBuilding Federated Multimodal AI Workflows with NVIDIA FLARE
• articleDeveloping Nemotron 3.5 Lightning NVFP4 with QAD Using NVIDIA Model Optimizer

• articleEnhancing Goodput in Large-Scale LLM Training with Nonuniform Tensor ParallelismMichelle Horton
• articleEfficient Memory Management for Large Language Model Serving with PagedAttentionWoosuk Kwon et al.
• videoWhy LLM Inference Is Memory-Bound (Julia Turc)Julia Turc

Checking sign-in…

Loading comments…