Log in List your tool
NVIDIA logo NVIDIA AI Computing Platforms View tool profile

NVIDIA updates

The full release history for NVIDIA — every new feature, pricing change and integration, sourced from their official changelog and social media, then verified by our editors.

RSS feed
292Updates tracked
14.6Avg. updates / month
Feb 2025Added On
Jul 27Last updated
Sourced from NVIDIA's official changelog and social media, verified by our editors.

Showing 6 of 292 updates

Earlier

AI Innovation Update

NVIDIA Reports Nemotron 3 Ultra Leads Open Models in Agentic Chip Design

NVIDIA announced benchmark results showing Nemotron 3 Ultra excelling at agentic RTL (Register Transfer Level) chip design. In the evaluation, the model repeatedly generated RTL code, tested it in a simulator, analyzed failures, and refined the design through multiple iterations. Across nine categories of real-world chip design tasks, Nemotron 3 Ultra achieved an average 97.1% pass rate while using 6,629 tokens per iteration, making it the best-performing open model tested in both design accuracy and efficiency.

via Source
Company News

NVIDIA Unveils Open-Source Cosmos 3 Edge World Model for Physical AI

NVIDIA introduced Cosmos 3 Edge at SIGGRAPH 2026, an open-source 4-billion-parameter world model designed for robots and edge AI devices. Running in real time on platforms such as Jetson and RTX PRO, the model processes text, images, video, audio, and actions, and is paired with a 2-billion-parameter reasoning model for tasks like robotics, autonomous driving, and live video analysis. NVIDIA also released the model's weights and source code, enabling developers to fine-tune it for edge AI applications while emphasizing simulation-first development for physical AI.

via Source
Company News

NVIDIA Unveils SpatialClaw for Code-Driven Spatial Reasoning Agents

NVIDIA AI has introduced SpatialClaw, a training-free agent from NVIDIA Research that uses Python code as its action interface for complex visual and spatial reasoning tasks. Instead of relying on a fixed set of predefined tools, SpatialClaw writes and executes Python within a persistent kernel, allowing it to compose perception modules, inspect intermediate outputs, and iteratively refine strategies across multiple steps. Visual perception outputs become reusable variables that can be combined with scientific libraries such as NumPy and SciPy. According to NVIDIA, SpatialClaw outperforms a recent baseline agent by 11.2 points across 20 benchmarks and maintains strong performance across six different model backbones, highlighting the growing importance of code execution as a general-purpose interface for embodied and spatial AI agents.

via Source
Collaboration

NVIDIA and LG Expand Partnership With Planned AI Factory

NVIDIA and LG Group have announced plans to build an AI factory aimed at supporting next-generation technologies across robotics, autonomous driving, data centers, and GPU cloud services. The announcement follows a meeting between Jensen Huang and Koo Kwang-mo as the companies deepen their AI collaboration. The initiative highlights growing investment in dedicated AI infrastructure designed to power industrial automation and large-scale AI workloads. These developments provide valuable industry insights for organizations evaluating premium subscription AI infrastructure platforms and best rates for enterprise AI deployment and cloud computing services.

via Source
AI Innovation Update

NVIDIA and SK hynix Form Multi-Year AI Memory Partnership

NVIDIA and SK hynix have announced a multi-year technology partnership to co-develop next-generation memory technologies for AI infrastructure. The collaboration will support future NVIDIA platforms ranging from NVIDIA Vera Rubin to Jetson Thor, while also advancing semiconductor manufacturing through digital twins built with NVIDIA Omniverse. The companies plan to leverage NVIDIA’s CUDA-X and PhysicsNeMo technologies to accelerate chip design and production workflows. These developments provide valuable industry insights for organizations evaluating premium subscription AI infrastructure solutions and best rates for large-scale AI factory deployments.

via Source
AI Innovation Update

NVIDIA Launches Nemotron 3 Ultra for AI Agents

NVIDIA has released Nemotron 3 Ultra, a 550-billion-parameter Mixture-of-Experts model designed for long-running AI agents and complex workflows. The model features a large context window, faster inference performance, and lower operating costs compared with other open frontier models. These advancements provide valuable industry insights for developers and enterprises comparing premium subscription AI platforms and best rates for agentic AI deployment. Explore NVIDIA Nemotron 3 Ultra and its advanced agent capabilities today.

via Source
🔍

No updates match your filters

Try a different keyword or clear a filter to see more.

Disclaimer

The content on this website is written and reviewed by experts in the fields of Artificial Intelligence and Software. Additionally, we may incorporate public opinions sourced from various social media platforms to ensure a comprehensive perspective. Please note that the screenshots and images featured on this website are sourced from NVIDIA' website. We extend our gratitude and give full credit to NVIDIA for their valuable contributions. This page may include external affiliate links, which could earn us a commission if you decide to make a purchase through those links. However, the opinions expressed on this page are our own, and we do not accept payment for favorable reviews.