The latest AI news and updates of NVIDIA — AI computing company building GPUs, inference hardware, and developer platforms for training and deployment. Covering NVIDIA's latest launches, product updates, and research from the past 90 days.
NVIDIAJul 28NVIDIA’s Cosmos world foundation models have reached 10 million downloads on Hugging Face. This milestone reflects widespread adoption among researchers and engineers building physical AI systems, including robots and autonomous vehicles. The open model platform provides the foundation for real-time scene understanding, reasoning, and action generation in memory-constrained edge environments.
NVIDIAJul 27NVIDIA introduced the Open Secure AI Alliance to develop industry-wide AI security standards. As a founding contribution, NVIDIA released the Nvidia Labs Object-Oriented Agents (NOOA) framework. This open-source research project structures agents as Python classes, achieving state-of-the-art performance on software engineering, cybersecurity, and reasoning benchmarks while reducing token costs through a typed, object-oriented harness architecture.
NVIDIAJul 27NVIDIA tested its Nemotron 3 Ultra model on agentic RTL chip-design tasks using the ACE-RTL agent. The model achieved a 97.1% average pass rate across nine design categories on the CVDP benchmark. It maintained this accuracy while using 6,629 tokens per iteration, outperforming other open models in both pass rate and token efficiency.
NVIDIAJul 24NVIDIA launched ModelExpress to accelerate model weight distribution, reducing DeepSeek-V4 Pro startup time from 8 minutes to under 2 minutes. The service moves weights directly between GPUs using RDMA via NIXL, bypassing centralized broadcasts. It also reuses JIT-compiled kernel caches across replicas, further reducing latency for inference and RL post-training workflows.
NVIDIAJul 24NVIDIA’s Kaggle Grandmasters team secured 1st place in the 2026 NeuroGolf Championship and 2nd in the KDD Cup 2026 Data Agents Competition. The team used manager agents with specialized workers for NeuroGolf, while their KDD Cup entry utilized a purpose-built harness to narrow the performance gap between a frontier LLM and a smaller open model.
NVIDIAJul 23NVIDIA demonstrated a hosted reinforcement learning workflow using Prime Intellect Lab that improved Nemotron 3 Nano’s math task accuracy from 22% to 91%. The process costs under $5 and produces a downloadable LoRA adapter. This same customization workflow scales to the larger Nemotron 3 Super and Ultra models by updating a single configuration line.
NVIDIAJul 22NVIDIA analyzed results from its Nemotron Model Reasoning Challenge, where 5,000+ participants optimized reasoning workflows using Nemotron-3-Nano-30B models. The findings highlight five practical habits for building reliable reasoning systems, including using verifiable chain-of-thought traces, designing traces to fit token budgets, and separating reusable knowledge from live problem-solving. These techniques were validated on Google Cloud G4 VMs with Blackwell GPUs.
NVIDIAJul 22NVIDIA released 4-step Cosmos 3 Super models that generate images and video up to 25x faster than the original versions. These models now rank first for image-to-video and second for text-to-image on the Artificial Analysis open-weight leaderboards. The new checkpoints are available for use on Hugging Face.
NVIDIAJul 22Wistron opened its first U.S. manufacturing facility in Fort Worth, Texas, to produce the NVIDIA GB300 Grace Blackwell Ultra Superchip. The $700 million plant utilizes digital twin technology to optimize production and will eventually manufacture the Vera Rubin Superchip. This facility expands domestic capacity for assembling and testing advanced AI infrastructure.
NVIDIAJul 21NVIDIA tested its Nemotron 3 Ultra model on the International Mathematical Olympiad 2026 problems under strict competition constraints, including no internet or external tools. The model achieved a score of 30/42, surpassing the official 29-point gold medal threshold. This result demonstrates the model's mathematical reasoning capabilities in a controlled, time-limited environment.
NVIDIAJul 21NVIDIA GB300 NVL72 achieved a world record of 1,648 TFLOPs per GPU while pre-training the DeepSeek-V3 671B model. This performance represents a 3x increase over the previous-generation GB200 NVL72. The throughput gains result from hardware-software co-design and continuous optimizations across the Megatron-Core, TorchTitan, and JAX frameworks.
NVIDIAJul 20NVIDIA released Cosmos 3 Edge, an open 4-billion-parameter world model designed for on-device physical AI. The model combines autoregressive and diffusion transformer towers to enable real-time scene understanding, prediction, and action generation for robots and autonomous systems. It ranks first on VANTAGE-Bench for vision analytics and includes model weights, post-training recipes, and code on Hugging Face.
NVIDIAJul 20NVIDIA updated the Agent Toolkit with new Omniverse libraries, providing AI agents with tools for physical AI simulation. The release includes ovrtx for sensor simulation, ovphysx for GPU-accelerated physics, and CAD-to-SimReady skills for asset validation. These components automate scene inspection and asset preparation within existing 3D applications, with integration blueprints now available on GitHub.
NVIDIAJul 17NVIDIA’s Nemotron 3 Embed 8B and 1B models secured the first and second positions on the Long-Horizon Memory (LMEB) benchmark. The models achieved scores of 64.36 and 61.50 across 22 tasks. This benchmark evaluates retrieval accuracy in long-running conversations and memory-intensive tasks, providing a measure for agentic systems that require context retention across sessions.
NVIDIAJul 17NVIDIA updated its NeMo AutoModel library to support Hugging Face Diffusers, enabling distributed fine-tuning for image and video models. The integration provides ready-to-run full and LoRA recipes for models like FLUX.1-dev and Wan 2.1 without requiring checkpoint conversion. Sharding and parallelism configurations now extend directly to Diffusers models, scaling training across multiple GPUs without architecture-specific wiring.
NVIDIAJul 16NVIDIA released the Nemotron 3 Embed 8B model, which secured the #1 ranking on the RTEB retrieval accuracy benchmark. The open-weights model supports a 32k context window and multilingual retrieval, aiming to improve agentic context accuracy. NVIDIA also introduced high-efficiency 1B variants, including a Blackwell-optimized NVFP4 version, to support production-scale retrieval and agent memory workflows.
NVIDIAJul 16NVIDIA is partnering with Noetra and Japan’s METI to build the world’s first national AI infrastructure for physical AI. The factory, built on the NVIDIA DSX platform, features 27,500 Rubin GPUs and 13,750 Vera CPUs to deliver 140 megawatts of capacity. It supports Japan’s FRONTia Project, providing foundation models for robotics and industrial applications.
NVIDIAJul 16NVIDIA released DeepStream 9.1, introducing 13 agentic skills to automate video analytics pipeline development. The update features Multi-View 3D Tracking for consistent object IDs across camera networks and AutoMagicCalib for automated camera calibration. It supports JetPack 7.2 for Jetson Orin and Thor edge platforms, with all source code and reference applications now available on GitHub.
NVIDIAJul 15NVIDIA GEAR Lab introduced RoboTTT, a robot foundation model that scales visuomotor context to 8,000 timesteps—three orders of magnitude beyond current policies—without increasing inference latency. The model uses test-time training to compress history into fast weights, enabling one-shot imitation from human video, on-the-fly self-correction, and improved performance on long-horizon tasks.
NVIDIAJul 14NVIDIA automated the post-training of its Cosmos 3 Nano model using TAO agent skills and AutoML. The workflow increased traffic signal detection accuracy on the Woven Traffic Safety dataset from a 54.41% zero-shot baseline to 93.35% in under a day. The process handles data patching, baseline evaluation, and hyperparameter sweeps through natural language prompts.
NVIDIAJul 14NVIDIA demonstrated an autonomous coding agent that uses NeMo RL, NeMo Gym, and reusable agent skills to manage reinforcement learning research. The agent built a visual counting environment and trained a Qwen3-VL-2B-Instruct model, increasing accuracy from 25% to 96.9%. It also autonomously proposed a follow-up experiment, while the researcher maintained strategic oversight of the campaign.
NVIDIAJul 10NVIDIA Research released Flex-Forcing, a video generation method that trains a single model to switch between bidirectional diffusion and autoregressive generation at inference time. This framework allows selection of a generation approach based on specific compute budgets, balancing structural consistency with streaming speed. The project was recognized with a spotlight at ICML 2026.
NVIDIAJul 7NVIDIA Research introduced MOTIVE, a motion-centric data attribution framework that identifies training clips influencing temporal dynamics in video generation. By re-weighting gradients toward moving regions, the method enables curation of high-influence subsets. Fine-tuning on these subsets improves motion smoothness and dynamic degree, achieving a 74.1% human preference win rate against the base model.
NVIDIAJul 6NVIDIA AI highlights an ICML 2026 research paper that estimates the memorization capacity of GPT-style models at approximately 3.6 bits per parameter. This metric distinguishes unintended data memorization from model generalization, offering a quantitative approach to evaluate training data requirements, scaling laws, and privacy risks in large language models.
NVIDIAJul 1NVIDIA Research released Nemotron-Labs-TwoTower, a diffusion language model adapted from the 30B-parameter Nemotron-3-Nano-A3B. The architecture splits the model into a frozen context tower and a trainable denoiser tower, enabling parallel token generation. This approach retains 98.7% of the original model’s quality while delivering 2.42× faster wall-clock throughput. Code and weights are available on Hugging Face.
NVIDIAJul 1NVIDIA Research introduced Generative Pretrained Controllers (GPC), a framework that models motor skills as discrete tokens for transformer-based next-token prediction. Trained on 600+ hours of motion data, the controller runs in real-time within physics simulations and achieves a 99.98% success rate in reproducing motion. The framework allows fine-tuning a single pretrained controller to solve various new downstream tasks.
NVIDIAJun 30NVIDIA GEAR Lab introduced ASPIRE, a system that autonomously discovers reusable robot skills by writing and refining control programs. It compounds experience into a library of 90+ skills across 150+ tasks, achieving up to a 10x reduction in transfer learning tokens. The system enables sim-to-real and cross-embodiment transfer by shipping code-based know-how rather than neural network weights.
NVIDIAJun 30NVIDIA released TAO 7, adding agent skills that plug into coding agents to improve accuracy. The update includes LLM-guided AutoML that finds hyperparameter configurations up to 2x faster, plus local fine-tuning for HuggingFace CV and VLM models. Data-enhanced fine-tuning also allows agents to identify and fix model failures automatically.
NVIDIAJun 28NVIDIA built the Enterprise Inference Hub to manage AI model access for thousands of internal engineers. Using LiteLLM as a central gateway, the platform provides a single API for over 100 models across cloud, open-source, and internal services. The hub now processes trillions of tokens every week, centralizing authentication, monitoring, and cost management for all internal AI applications.
NVIDIAJun 24NVIDIA NeMo AutoModel now integrates HuggingFace Transformers v5, adding Expert Parallelism, DeepEP, and TransformerEngine kernels. This update delivers 3.4–3.7x higher training throughput and 29–32% lower peak GPU memory usage for Mixture-of-Experts models. The framework maintains API compatibility, enabling these performance gains through a single import line change without requiring additional code rewrites.
NVIDIAJun 23NVIDIA integrated DFlash, an open-source block diffusion model, to accelerate inference on Blackwell GPUs by up to 15x. By drafting entire token blocks in parallel rather than sequentially, DFlash maintains interactivity targets while boosting throughput. The update is available as a drop-in integration for SGLang, TensorRT-LLM, and vLLM, with 20 model checkpoints released on Hugging Face.
NVIDIAJun 23NVIDIA launched the BioNeMo Agent Toolkit to turn scientific workflows into repeatable, agent-executable tasks. By integrating Nemotron, NemoClaw, and OpenShell, the toolkit provides agents with domain-specific skills for biology, chemistry, and drug discovery. It enables agents to synthesize scientific knowledge, call models, and execute complex experiments, accelerating discovery across life sciences research.
NVIDIAJun 22NVIDIA Research released ArtiFixer, an open autoregressive model that fills in missing geometry in 3D scene reconstructions. The two-stage pipeline uses a bidirectional generative model and causal distillation to generate consistent novel views in unobserved areas. It outperforms existing baselines by 1–3 dB PSNR and provides a method for generating plausible reconstructions where other approaches fail.
NVIDIAJun 22NVIDIA launched Halos for Robotics, an open, full-stack safety system for physical AI. The platform unifies AI compute, sensor connectivity, and safety applications, including the Halos Outside-In Safety Blueprint. Agility Robotics is the first partner to integrate the system into its humanoid robots. Halos Core for IGX is available in early access, with the safety blueprint open-sourced on GitHub.
NVIDIAJun 17NVIDIA GEAR Lab’s ENPIRE system enables eight Codex-controlled robots to autonomously self-improve manipulation policies overnight. The framework implements a two-layer safety harness, freezes reward classifiers to prevent objective gaming, and uses live telemetry metrics—including Mean Robot Utilization and Mean Token Utilization—to maintain resource awareness during unattended research.
NVIDIAJun 16NVIDIA Research released SpatialClaw, a training-free agent framework that uses Python code as its action interface for spatial reasoning tasks. The agent composes perception modules and revises strategies across steps, achieving a 59.9% average accuracy across 20 benchmarks. This approach outperforms the prior best spatial agent by 11.2 points and functions consistently across six different VLM backbones.
NVIDIAJun 16NVIDIA’s GEAR lab introduced ENPIRE, a framework that gives Codex agents autonomous control of a robot fleet to self-improve manipulation policies. The system solves high-precision tasks like zip-tie tying and GPU installation. Researchers discovered that eight robots exploring in parallel improve significantly faster than smaller teams, and the lab plans to open-source the entire system.
NVIDIAJun 15NVIDIA Research released MotionBricks, a single open model trained on 350,000+ motion clips that generates real-time character animation at 15,000 FPS. The framework eliminates hand-crafted transitions and fine-tuning, applying zero-shot to both animation and robotics. Preview code is available now in the GR00T Whole-Body Control repository, with a full release targeted for approximately one month from now.
NVIDIAJun 13NVIDIA and Abridge are collaborating to build a specialized foundation model for clinical conversations. The model is designed to understand doctor-patient interactions, generate structured clinical notes, and support decision-making within existing medical workflows. Abridge CEO Shivdev Rao emphasizes that this specialized training is necessary to ensure clinical intelligence is properly shaped and evaluated against real-world conditions.
NVIDIAJun 13NVIDIA now provides a free GPU-accelerated endpoint for MiniMax M3, a 428B-parameter Mixture-of-Experts model. The endpoint exposes M3's 1-million-token context window and native multimodal reasoning across text, image, and video, available for prototyping and evaluation on the NVIDIA build platform.
NVIDIAJun 13NVIDIA announced day-one support for Google DeepMind's DiffusionGemma, an experimental model that generates 256 tokens in parallel per step. BF16 and NVFP4 checkpoints are available on Hugging Face, alongside free GPU-accelerated endpoints and vLLM deployment. The model delivers over 150 tokens per second on DGX Spark and up to 1,000 on a single H100 GPU.
NVIDIAJun 9NVIDIA trained Llama 3 8B and 405B models on its Blackwell platform using NVFP4 precision. This achieved a 1.31–1.73x speedup compared to FP8 precision, with no loss in accuracy. The update demonstrates how specialized hardware and precision formats can significantly boost the efficiency of large language model development.
NVIDIAJun 8NAVER is expanding its AI infrastructure in Korea by building a full-stack NVIDIA AI factory, powered by the NVIDIA DSX platform. This initiative includes plans to scale its GAK Sejong data center from 55 megawatts to gigawatt capacity, addressing the surging demand for AI compute. The collaboration aims to provide sovereign AI capabilities for enterprises, industries, and government customers.
NVIDIAJun 8NVIDIA and Doosan Group announced an expanded collaboration to advance physical AI and AI factory infrastructure. This partnership will explore robotics, AI factory power solutions, and advanced electronics materials for next-generation data center systems. The initiative aims to integrate AI across industrial automation and energy sectors.
NVIDIAJun 8NVIDIA and LG Group announced plans to build an AI factory. This facility will provide accelerated computing infrastructure to support LG's AI-driven businesses, spanning robotics, autonomous driving, and GPU cloud services, by integrating NVIDIA's full-stack AI platform.
NVIDIAJun 6NVIDIA Research's PixelDiT, a single-stage image generation model, was selected as a best paper finalist at CVPR 2026. It removes the autoencoder step common in most image generation models, learning the diffusion process directly in pixel space to preserve fine details and achieve state-of-the-art performance among pixel-space generative models. This approach addresses a key limitation in image quality by eliminating a lossy compression step.
NVIDIAJun 5NVIDIA's Nemotron Coalition, which develops open frontier AI models, has expanded its membership to include hcompany_ai, NousResearch, and PrimeIntellect. This growth continues the collaborative effort, with Prime Intellect specifically contributing to scaling agentic capabilities.
NVIDIAJun 5NVIDIA has released Nemotron 3 Ultra, an open model designed for long-running AI agents, and provided a tutorial for its setup and demonstrations. This model aims to make complex, multi-step agentic workflows faster and more cost-effective by delivering high throughput and efficient reasoning.
NVIDIAJun 5NVIDIA's NitroGen foundation model for generalist gaming agents received a CVPR Best Paper Honorable Mention. This recognition marks progress toward AI agents that master physics across both real-world and simulated environments, a step toward more adaptable, general-purpose embodied AI.
NVIDIAJun 4NVIDIA has shipped Nemotron 3 Ultra, a 550B Mixture-of-Experts (MoE) open model designed for long-running AI agents. This model delivers 5x faster inference and up to 30% lower cost for complex agentic tasks compared to other open frontier models, aiming to make autonomous workflows more efficient and accessible.
NVIDIAJun 4NVIDIA updated its secure agent runtime, OpenShell, to include support for Google Vertex AI as an inference provider. The release improves security transparency through profile-backed policy visibility and fixes critical sandbox behaviors for GPU and container environments. This expansion allows enterprise agents to run securely while accessing models across different cloud ecosystems.
NVIDIAJun 2NVIDIA introduced a single-command installer for NemoClaw on DGX Spark to accelerate the deployment of on-premise AI agents. The update delivers a 2.6x performance boost for Qwen3.6 models and automates multi-node clustering. This allows organizations to run complex, long-running agents locally while maintaining data privacy and eliminating cloud costs.
NVIDIAJun 2NVIDIA introduced the Factory Operations Blueprint (FOX), a reference design for creating centralized AI agents that orchestrate industrial workflows. The system allows manufacturers to connect real-time machine data with specialized agents to automate root-cause analysis and production optimization.
NVIDIAJun 1NVIDIA released Cosmos 3, an open-weights foundation model family designed for physical AI applications like robotics and autonomous driving. By unifying vision reasoning and world simulation into a single architecture, the model allows developers to build autonomous systems that understand physical laws and predict future states within one workflow.
NVIDIAJun 1NVIDIA launched DSX MaxLPS and DSX OS to help infrastructure builders optimize power consumption and manage large-scale AI data centers. These open-source tools allow operators to run more hardware within existing power limits while automating complex facility operations.
NVIDIAJun 1NVIDIA announced new DOCA security features for the Vera BlueField-4 STX platform to protect autonomous agents and context memory. By moving threat detection into silicon, the system achieves 1,000x faster response times to secure the data path in real-time AI environments.
NVIDIAJun 1NVIDIA unveiled the RTX Spark superchip, a 1-petaflop processor combining Blackwell and Grace architectures with 128GB of unified memory for Windows PCs. The platform introduces a native security layer and dedicated runtime to allow autonomous agents to execute complex workflows privately on local hardware.
NVIDIAJun 1NVIDIA announced the DGX Station for Windows, a deskside supercomputer powered by the GB300 Grace Blackwell chip. It allows enterprises to run frontier-class AI models and autonomous agents locally within their existing Windows infrastructure. This shift bridges the gap between high-performance Linux data centers and the Windows applications where professional work actually happens.
NVIDIAJun 1NVIDIA introduced Alpamayo 2 Super, a 32-billion-parameter reasoning model that allows autonomous vehicles to plan and explain complex driving decisions. The release includes simulation and reinforcement learning tools to help developers scale robotaxi fleets safely.
NVIDIAJun 1NVIDIA announced the Isaac GR00T Reference Humanoid Robot, an open platform combining Unitree hardware, Sharpa dexterous hands, and Jetson Thor compute. The design aims to solve fragmentation in robotics research by providing a standardized hardware and software stack for developing general-purpose humanoid behaviors.
Frequently asked questions