News Hub

FLUX 3 Image Klein achieves faster image generation on NVIDIA Blackwell GPUs.

Reshard Collective accelerates reinforcement learning weight refit by parallel inference and training.

NVIDIA co-design methods optimize high bandwidth memory usage in AI systems to reduce redundant data copies.

NVIDIA uses AI agents to help site reliability engineers diagnose GPU cluster performance issues across compute, network, hardware, and scheduling systems.

NVIDIA CUDA 13.5 reaches general availability with support for NVIDIA Vera Rubin GPU architecture.

NVIDIA NeMo converts agent failures into tests and evaluations to improve prompts, tools, models, and code.

NVIDIA addresses distributed AI training failures through automated workload recovery to prevent stalled training ranks.

Kubernetes configuration for NVIDIA Vera Rubin GPU hardware requires coordinating drivers, kernels, runtimes, networking, operators, and schedulers.

NVIDIA Cosmos is a platform for building world models for interactive simulation.

NVIDIA Dynamo is a distributed inference framework that coordinates model serving across GPU workers on Kubernetes.

Developers can deploy AI models at the edge for applications like driver monitoring, robot navigation, and fleet coordination.

NVIDIA NemoClaw enables building specialized agents with controlled runtime, policies, optimization, and observability.

NVIDIA GPUDirect Storage accelerates GPU storage access by providing efficient data paths between GPUs and storage.

NVIDIA CUDA 13.4 enables GPU memory sharing in Kubernetes to serve multiple AI workloads on single GPUs.

NVIDIA RTX Spark is a Windows PC platform for AI development powered by Blackwell RTX GPU and Grace CPU.

Fine-tuning NVIDIA Nemotron enables language models to learn local rules and cultural knowledge.

Training large-scale AI models on thousands of GPUs is a high-stakes operation where monitoring and automation must be built-in to optimize cluster resiliency... Training large-scale AI models on thousands of GPUs is a high-stakes operation where monitoring and automation must be built-in to optimize cluster resiliency and uptime at scale. This concept isn’t new to data center operators who have measured mean-time-to-repair (MTTR) and mean-time-between-failure (MTBF) long before AI factories existed. Now the stakes are higher. Modern AI factories run… Source

NVIDIA TensorRT 11.1 includes Global Performance Tuner for automatic accuracy-aware performance tuning.

Neural network deployment involves compilation decisions with multiple legitimate optimization routes.

NVIDIA GPUDirect Storage enables reading large datasets directly into GPU memory in Python.