Daily AI news, sourcedOfficial news, first rumor reports and substantive progress. Each entry keeps the information published at that time.
- NVIDIA plans RTX Spark laptop launch events in six cities
NVIDIA announced launch events for its RTX Spark Windows laptops in Beijing, Berlin, Istanbul, Seoul, San Francisco and Taipei, with demos, swag and a chance to win an ASUS ProArt P14 RTX Spark laptop signed by Jensen Huang. NVIDIA says event details are coming soon.
- NVIDIA contributes Vera Rubin rack and compute tray designs to OCP and open-sources new DSX AI factory software
At the Open Compute Project Global Summit, NVIDIA contributed Vera Rubin rack and compute tray designs through its open MGX architecture, including the MGX 1.2 spec with blind-mate liquid cooling and support for racks up to 250kW, plus a Vera CPU rack design for agent tool calling and sandboxed execution. It also released open-source DSX OS components for network configuration, cluster stress-testing, rack management and bare-metal provisioning. NVIDIA joined AMD, Google and Meta on the Open Data Center Specification v0.7, and Google, Meta, Microsoft and NVIDIA published an OCP Grid Resilience Specification.
- NVIDIA reportedly ends new RTX 5090 GB202 supply for GeForce, reserves chips for RTX PRO, per VideoCardz
VideoCardz reports leaker MEGAsizeGPU saying NVIDIA will no longer allocate new GB202 silicon to GeForce RTX 5090 production and will reserve future GB202 supply for RTX PRO workstation cards. A separate factory-notice claim says an RTX 5080 24GB would become the top GeForce card; NVIDIA has not confirmed either report.
- NVIDIA publishes a GR00T N1.7 deployment checkpoint for an Agile One S SSD-placement robot task
NVIDIA published a GR00T N1.7 deployment model on Hugging Face for picking up and placing an SSD with three cameras, republishing the original deployment assets without retraining. The repository includes five ONNX graphs and two TensorRT BF16 engines, and NVIDIA notes that file availability is not a robot deployment or safety qualification.
- NVIDIA shows developers building Omniverse sims with frontier AI agents such as GPT-6 Astra
An NVIDIA Omniverse blog walks through projects where developers direct frontier AI agents with natural-language instructions while Omniverse libraries supply GPU physics, rendering, and sensor simulation. Examples include a warehouse humanoid simulator wired with Astra and SimReady assets, a Zero to Alpamayo autonomous-driving test bed on a San Francisco Market Street scene, sensor-matched digital twins refined with Astra and Claude Fable 5, Robo Olympics Unitree G1 sports trials in Newton/Warp, CAD-informed robotic disassembly in Isaac Sim, a browser ISS OpenUSD viewer with telemetry, and stereo-capture rooms turned into editable OpenUSD studios.
- NVIDIA commits $1 billion over five years to US science and super-intelligence research
NVIDIA announced commitments valued at $1 billion over the next five years to advance US scientific capacity in quantum computing, healthcare and energy security, unveiled at a Washington 'Science: A New Golden Age' event tied to the Genesis Mission. The company said the package supports higher-education research, quantum leadership and cloud capacity for government missions, building on two decades of national-lab work including DOE systems at Argonne and Los Alamos.
- NVIDIA AI posts how H Company optimizes VLM serving for computer-use agents with Dynamo
NVIDIA AI shared a technical write-up on how H Company optimizes vision-language model serving for computer-use agents using NVIDIA Dynamo.
- NVIDIA congratulates CEO Jensen Huang on the National Medal of Science for GPU computing
NVIDIA said CEO Jensen Huang received the National Medal of Science for advancing GPU computing to power scientific breakthroughs. The post also congratulated fellow honorees Elon Musk, Lisa Su, Michael Dell, Satya Nadella and Sergey Brin.
- NVIDIA AI streams a walkthrough of what's new in CUDA 13.4
NVIDIA AI hosted a live broadcast on X covering what's new in CUDA 13.4.
- NVIDIA researchers unveil PivotOPD to help AI agents recover from early mistakes
NVIDIA AI said its researchers built PivotOPD to teach agents how to avoid early mistakes that send a task in the wrong direction, and how to recover when they happen. During training, a teacher model shows the agent a better action and how to get back on track over the next few steps. The paper and a demo are on NVIDIA Research's site.
- NVIDIA fine-tunes Nemotron to gold-level results at both IOI and IMO
In a Hugging Face blog post, NVIDIA says it started from Nemotron 3 and used supervised fine-tuning, reinforcement learning and test-time compute to build specialist models reaching gold-level results at the International Olympiad in Informatics and the International Mathematical Olympiad, and is releasing the open models, data and recipes.
- NVIDIA says telecom operators are building AI strategies on open models
An NVIDIA blog post, drawing on its State of AI in Telecommunications report, says telecom operators are basing AI strategies on open models for trust, control and customization, from autonomous networks to customer care.
- NVIDIA launches DGX Spark 64GB for local AI development
NVIDIA announced the DGX Spark 64GB, giving developers more ways to build and scale local AI, per the company's blog.
- NVIDIA: build visual AI agents from a prompt with Cosmos and VSS 3.3
NVIDIA announced that developers can build visual AI agents from a single prompt using NVIDIA Cosmos together with VSS 3.3, per the company's AI account.
- NVIDIA research has AI agents rewrite Kimi Delta Attention GPU kernels
NVIDIA research had AI agents directly rewrite GPU kernels for Kimi Delta Attention, the attention variant used in Moonshot AI's Kimi-Linear models.
- NVIDIA highlights Korean team's top-5 open model score on Artificial Analysis
NVIDIA's AI account highlighted how a small Korean team hit a top-5 AA II score in an open Nemotron Labs model.
- Nvidia pairs OpenShell agent software with Sentry hardware watchdog, per The Decoder
The Decoder reports Nvidia is combining its OpenShell agent software with Sentry, a hardware watchdog meant to isolate breakout AI agents within milliseconds, to create the Open Agent Safety Platform.
- Nvidia's Nemotron-Labs-3 Competitive Coding 550B model page live on Hugging Face
A page for Nvidia's NVIDIA-Nemotron-Labs-3-Competitive-Coding-550B-A55B-NVFP4, an open competitive-coding model, is live on Hugging Face.