NVIDIA Developer

Starting the day seeing NVIDIA Nemotron 3 Nano trending at #1 on ‪@HuggingFace‬🤗

Huge thank you to our developer community! Tell us what you are building with it.

➡️nvda.ws/3Mzz3V1

11 hours ago | [YT] | 87

NVIDIA Developer

✨ Meet our new open family of models: NVIDIA Nemotron 3 ✨
Open in weights, data, tools, and training, Nemotron 3 is built for multi-agent apps and features:

• An efficient hybrid Mamba‑Transformer MoE architecture
• 1M token context for long-term memory and improved reasoning
• Multi‑environment reinforcement learning via NeMo Gym for advanced skill adaptation

Plus NVFP4 pre-training, latent MoE, 1T tokens of data, and more.

📗Read the details in our tech blog: nvda.ws/4565FMe

🤗 Try the model on Hugging Face: huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A…

1 day ago | [YT] | 127

NVIDIA Developer

🤖 Don't miss our essential live session for robotics developers ready to leverage NVIDIA Isaac Sim on the powerful DGX Spark platform.
📆 Friday December 12, at 11 a.m. Pacific
📽️ https://youtube.com/watch?v=34PJYWPcqK4

We'll guide you through the initial steps to get your high-fidelity simulation environment up and running on the aarch64 architecture, with live Q&A.

Whether you're starting a new project or migrating an existing one, tune in to jump straight into developing your next-generation AI-powered robots.

#SparkSomethingBig

1 week ago | [YT] | 229

NVIDIA Developer

Learn how EPRI is using NVIDIA DGX Spark to push the boundaries of AI innovation. 🎥 Catch the full conversation: youtube.com/live/7U-NvEfKVfc?si=VYWdMobAYddSsKdc

“Our R&D team uses DGX Spark to run advanced open-weight models, including get-oss-120b and Qwen-3-Embedding-8B—powering a secure, AI-driven document research assistant.

This compact yet powerful system enables small teams to achieve high-quality results in the enterprise infrastructure, making custom AI workflows more accessible and compliant.”

– Matt Corbett, Director, Information Technology Enterprise Infrastructure & Networking, EPRI

#SparkSomethingBig ✨

1 week ago | [YT] | 229

NVIDIA Developer

CUDA-X is reshaping how scientists and engineers solve the world’s hardest problems, from data science pipelines to quantum simulations and semiconductor manufacturing, Catch the NVIDIA GTC D.C. session replay https://youtu.be/jbpzSnQxC5c to learn how 900+ domain-specific CUDA-X libraries help turn ideas into discoveries faster, with Jeff Larkin walking through real workloads and use cases.

2 weeks ago | [YT] | 187

NVIDIA Developer

Curious how open-source AI is enabling innovation?

Watch the NVIDIA GTC D.C. replay and hear from leaders like Bryan Catanzaro and Joseph Jacks on how open models foster collaboration, research, and progress across industries.

🎬 https://youtu.be/VqIc2LJzZG0

Skip the Carousel

1 2

2 weeks ago (edited) | [YT] | 88

NVIDIA Developer

CUDA 13.0—New Features and Beyond | NVIDIA GTC D.C.

Watch the replay -- available now -- to explore the latest features and what’s next in GPU computing. 🎥 https://youtu.be/_vL8uz7kf0M

2 weeks ago | [YT] | 190

NVIDIA Developer

✨ New DGX Spark Livestream on Friday: Process Text for GraphRAG With Up to 70B LLM ✨
📆 November 21 at 11 a.m. Pacific.
📹 Watch here: https://www.youtube.com/watch?v=uQtzj...

See NVIDIA DGX Spark in action for local GraphRAG and graph analytics. We will showcase how this demo runs entirely on DGX Spark’s 128GB unified memory, extracting nodes and relationships from local biomedical research papers to build a context-aware Knowledge Graph in ArangoDB.

3 weeks ago | [YT] | 232

NVIDIA Developer

📣 For NVIDIA DGX Spark Owners 📣 **New DGX Spark Software Updates for Performance, Stability, and Creator Workflow Improvements**

Software updates will begin rolling out to all systems through repositories starting today. These updates will also apply to partner systems based on the NVIDIA GB10 Grace Blackwell Superchip.

🧑‍💻✨ You can initiate the updates either through the DGX Dashboard or through the usual command line instructions to pull updates and upgrades.

This release is part of our commitment to continually enhance the DGX Spark experience based on real-world user feedback – bringing meaningful performance, stability, and workflow improvements across the OS, GPU stack, JupyterLab, and connectivity. If you’re in the middle of testing or planning deeper workflows, this update will materially improve your experience.

What’s new:

✅ New DGX OS Kernel (Ubuntu 6.14 HWE stack): Brings ongoing performance gains, tighter stability, broader hardware compatibility, and the latest security updates for a more secure operating environment.

✅ Updated JupyterLab Environment: Now running CUDA 13.0.2 and the latest PyTorch stack, enabling users to work with updated frameworks immediately without extra downloads.

✅ Accurate Memory Reporting: DGX Dashboard now aligns with CUDA unified memory guidance.

✅ Improved Peripheral Interoperability: Better compatibility with USB-C devices, monitors, Bluetooth peripherals, and Wi-Fi access points.

✅ Stable Diffusion XL Fix: Example workflow now runs end-to-end inside JupyterLab.

✅ Recovery Image Reliability: Installs correctly on macOS and when multiple external USB drives are connected.

✅ Keyboard Accessibility Improvements: Smoother and more complete Out-of-Box-Experience navigation.

In addition, we are actively working with third-party ecosystem partners to bring their software to DGX Spark. For example, we have worked with Llama.cpp to improve memory management on unified memory architecture-based systems to ensure accurate reporting of available resources.

If you’re actively benchmarking, generating content, or comparing Spark to OEM systems, we’d love to hear about your experience.

🙌 Thank you again for taking the time to work with the world’s first AI supercomputer that sits on your desk. We can’t wait to see what you build with it. #SparkSomethingBig ✨

More info here ➡️ docs.nvidia.com/dgx/dgx-spark/os-and-component-upd…

4 weeks ago | [YT] | 191

NVIDIA Developer

⏰ Livestream starting at 11 a.m. PST today. We’ll demonstrate how to use 2 DGX Spark systems as a desktop cluster to accelerate cuPyNumeric locally. ➡️ youtube.com/live/ghysPxEUZwU

NVIDIA cuPyNumeric, a multi-node, multi-GPU NumPy-compatible library, is ideal for domain scientists and researchers to scale their Python and NumPy program to multi-GPU multi–node without having to write complex parallel computing code like MPI.

#SparkSomethingBig ✨

1 month ago | [YT] | 185