⚡ NVIDIA SPINGE SULL’OPEN AI: SLURM NELL’ECOSISTEMA E NASCE NEMOTRON 3 ⚡
NVIDIA is aggressively strengthening its "open" AI strategy with a dual move: the acquisition of SchedMD, the developer of Slurm, and the launch of the new family of open-source models, Nemotron 3.
Slurm is the most widely used open-source workload manager in supercomputers and AI clusters for scheduling training and inference jobs, and NVIDIA has promised that it will remain vendor-neutral and open source, while integrating it more deeply with its GPUs and Blackwell architectures.
This means that researchers and companies will be able to better orchestrate huge heterogeneous clusters, optimizing throughput and costs for the training of LLMs and generative workloads.
In parallel, Nemotron 3 arrives, a new family of open models designed for advanced AI agents and multi-agent systems.
The line includes the Nano, Super, and Ultra variants, based on a hybrid Mixture-of-Experts architecture that aims to combine high accuracy, long-horizon reasoning, and lower inference costs, ideal for agents cooperating with each other in complex scenarios.
Nemotron 3 is released with training datasets, RL libraries (NeMo Gym, NeMo RL), and open-source evaluators on GitHub and Hugging Face, positioning NVIDIA as a supplier of a complete stack: GPUs, scheduler (Slurm), models (Nemotron), and tools for reinforcement learning.
For the market, the message is clear: NVIDIA wants to dominate not only AI hardware but also the open software infrastructure on which the next agents and "Physical AI" systems will run.
#breakingnews #NVIDIA #ArtificialInteligence