LAS VEGAS, NV (CES 2026), Jan 9, 2026 – NVIDIA has released open models, frameworks, and AI infrastructure focused on physical AI, which applies AI to machines operating in real-world environments. The updates streamline robot development by supporting simulation, training, and deployment within a single workflow. The release highlights partner robotics for manufacturing, logistics, and healthcare. The tools support robots with multi-task learning and environment adaptation.
“The ChatGPT moment for robotics is here. Breakthroughs in physical AI – models that understand the real world, reason and plan actions – are unlocking entirely new applications,” said Jensen Huang, founder and CEO of NVIDIA. “NVIDIA’s full stack of Jetson robotics processors, CUDA, Omniverse and open physical AI models empowers our global ecosystem of partners to transform industries with AI-driven robotics.”
New Open Models Advance Robot Learning and Reasoning
Developing robots capable of performing multiple tasks requires significant computing resources and model training. NVIDIA is providing open foundation models to reduce pretraining requirements and support development of AI-based robots and autonomous machines.
New models available on Hugging Face include:
- NVIDIA Cosmos Transfer 2.5 and NVIDIA Cosmos Predict 2.5 – open and customizable models that enable physically based synthetic data generation and robot policy evaluation in simulation for physical AI.
- NVIDIA Cosmos Reason 2, an open reasoning vision language model (VLM) that enables machines to see, understand and act in the physical world like humans.
- NVIDIA Isaac GR00T N1.6, an open reasoning vision language action (VLA) model, built for humanoid robots, that uses NVIDIA Cosmos Reason for reasoning and understanding.
Franka Robotics, NEURA Robotics and Humanoid are using GR00T-enabled workflows to simulate, train and validate new behaviors for robots. Salesforce is using Agentforce, Cosmos Reason and the NVIDIA Blueprint for video search and summarization to analyze video footage captured by its robots and reduce incident resolution times by 2x.
LEM Surgical is using NVIDIA Isaac for Healthcare and Cosmos Transfer to train the autonomous arms of its Dynamis surgical robot, powered by Jetson AGX Thor and Holoscan. XRlabs is using Thor and Isaac for Healthcare to enable surgical scopes, starting with exoscopes, to guide surgeons.
New Open-Source Simulation and Compute Frameworks for Robotics Development
NVIDIA has released open-source simulation frameworks on GitHub to address fragmentation in robotics training workflows. The tools reduce manual benchmarking and simplify end-to-end pipelines that span multiple compute resources. The release aims to support the transition from research environments to robotics deployments.
NVIDIA has released Isaac Lab-Arena, an open-source framework on GitHub for evaluating and benchmarking robot policies in simulation. Developed with Lightwheel, Isaac Lab-Arena integrates with Libero and Robocasa to standardize tests and validate robot skills prior to deployment on physical hardware.
NVIDIA has introduced OSMO, a cloud-native orchestration framework for robotics development. OSMO coordinates synthetic data generation, model training, and software-in-the-loop testing across different compute environments and mixed cloud compute resources. OSMO is used by robot developers like Hexagon Robotics, and integrated into the Microsoft Azure Robotics Accelerator toolchain.
NVIDIA and Hugging Face Accelerate Open-Source Physical AI Development
Robotics is now the fastest-growing category on Hugging Face, where NVIDIA’s open models and datasets lead downloads among a surging open-source community.
NVIDIA and Hugging Face are integrating open-source Isaac and GR00T technologies into the LeRobot. The work follows growth in robotics usage on Hugging Face, where NVIDIA-provided models and datasets lead downloads. The integration links NVIDIA’s 2 million robotics developers with Hugging Face’s global community of 13 million AI builders.
The LeRobot library includes GR00T N models and Isaac Lab-Arena for robot policy tuning and evaluation. Hugging Face has made its Reachy 2 humanoid compatible with NVIDIA Jetson Thor, enabling execution of VLA models, including GR00T N1.6. The Reachy Mini tabletop robot is also interoperable with NVIDIA DGX Spark to support local experimentation with LLMs, vision, and voice models.
Humanoid Robot Developers Adopt NVIDIA Jetson Thor
Humanoid robots developers are adopting NVIDIA Jetson Thor to support onboard reasoning and control. At CES, NEURA Robotics and Richtech Robotics presented new humanoids built on the platform (Porsche-designed Gen 3 humanoid and a mobile humanoid Dex ), including systems designed for dexterous manipulation and industrial navigation. AGIBOT introduced humanoids targeting both industrial and consumer sectors, along with the Genie Sim 3.0, a simulation platform integrated with Isaac Sim. LG Electronics also unveiled a home robot designed to perform indoor household tasks.
Bringing Physical AI to the Industrial Edge
NVIDIA has introduced the NVIDIA Jetson T4000 module to extend the Blackwell architecture into autonomous machines and robotics platforms. The module offers 1,200 FP4 TFLOPS and 64GB of memory, delivering 4x the performance of the previous generation. Designed for power-constrained systems, it operates within a configurable 70-watt envelope and is priced at $1,999 at 1,000-unit volume.
NVIDIA plans to release IGX Thor to support AI-driven robotics at the industrial edge. The platform combines high-performance AI computing, enterprise software support, and functional safety. Archer is adopting IGX Thor to deploy AI across aviation, focusing on safety, airspace integration, and future autonomy.
Hardware partners are delivering Thor-powered platforms for edge AI, robotics, and embedded systems, including offerings from Advantech, ADLINK, Aetina, AVerMedia, Connect Tech, EverFocus, ForeCR, Lanner, RealTimes, Syslogic, Vecow and YUAN. Caterpillar is expanding to apply AI and autonomous technologies to construction and mining operations.
Source: NVIDIA
About NVIDIA
NVIDIA, founded in 1993 and headquartered in Santa Clara, CA, designs and manufactures graphics processing units, systems on chips, networking hardware, and AI intelligence software such as CUDA. Its products serve industries including gaming, data centers, autonomous vehicles, professional visualization, robotics, health care, and energy. The company introduced the GPU in 1999 and later expanded into accelerated computing and AI infrastructure. In gaming, its GPUs support high-performance rendering, while in AI and high-performance computing, its systems provide the infrastructure for training and deploying large-scale models. NVIDIA also develops tools for robotics and autonomous driving.