Tag: nvidia
-
NVIDIA Nemotron 3 Super: 120B Hybrid Model Activates Only 12B Parameters for Agentic AI
On March 11, 2026, NVIDIA released Nemotron 3 Super — a 120-billion-parameter open-weight model that activates only 12 billion parameters per token, delivering frontier-class reasoning at a fraction of the compute cost. Designed for multi-agent AI systems, the model combines three distinct architectures into a single hybrid backbone and ships with a native 1-million-token context […]
-
NVIDIA Open-Sources SONIC: A Foundation Model for Humanoid Whole-Body Control
Key points NVIDIA has open-sourced SONIC, a 42-million-parameter humanoid whole-body control foundation model trained on over 100 million frames of motion-capture data, released on February 20, 2026. SONIC scales model capacity from 1.2 million to 42 million parameters, using 100M+ frames of motion capture and 9,000 GPU hours across 128 GPUs over 3 days. Its […]
-
🎵 Music Flamingo (MF) — Advanced Music Understanding with AI
Music Flamingo is a cutting-edge research project by NVIDIA’s Applied Deep Learning Research (ADLR) group that advances how AI systems understand music audio — not just speech. (NVIDIA) Key points Music Flamingo is a research project by NVIDIA’s Applied Deep Learning Research group that advances how AI systems understand music audio, not just speech. It […]
-
Advancing LLM Training: Introducing NVFP4 for Efficient Pretraining
Key points The study presents NVFP4, NVIDIA 4-bit floating point precision, as a way to improve large language model training efficiency without compromising model quality. The article states that FP4 training poses critical challenges including training stability for large-scale models, convergence issues with long token sequences, and implementation complexity in maintaining accuracy. The NVFP4 framework […]
-
Nvidia Introduces Lower-Cost Blackwell AI Chip for China Amid Export Restrictions
Content In response to stringent U.S. export controls, Nvidia is set to launch a more affordable AI graphics processing unit (GPU) tailored for the Chinese market. This new chip, based on Nvidia’s latest Blackwell architecture, is expected to be priced between $6,500 and $8,000, significantly lower than the $10,000–$12,000 price tag of the previously available […]
-
NVIDIA reportedly prepares for a ban on the GeForce RTX 5090D in China
Key points NVIDIA is reportedly preparing for a ban on the GeForce RTX 5090D in China. NVIDIA is reportedly preparing for a ban on the GeForce RTX 5090D in China.
-
Vietnam signs agreement with NVIDIA for AI research and data centers
Key points Vietnam signed an agreement with NVIDIA to establish AI research and data centers. Vietnam has signed an agreement with NVIDIA to establish AI research and data centers.
-
GR00T for Autonomous Robots
Key points NVIDIA’s GEAR Lab presented GR00T, which the article describes as a foundation model specially designed for autonomous robots. The article says GR00T was released as part of the GTC keynote from NVIDIA. NVIDIA GEAR Lab presents GR00T, a foundation modal specially designed for autonomous robots. Released as part of the GTC Keynote from […]
-
NVIDIA unveils Blackwell GPU and new AI innovations at GTC 2024
Key points NVIDIA announced the Blackwell GPU with 208 billion transistors, enabling large AI models and high-performance inference. It introduced the NeMo Megatron framework, Inference Microservices (NIMs), and Omniverse Cloud for digital twins. Partnerships with AWS, Google, Microsoft and Siemens were announced to accelerate AI across industries. NVIDIA announced the Blackwell GPU with 208 billion […]