NVIDIA Alpamayo family graphic illustrationNVIDIA official website
At CES 2026, NVIDIA unveiled a major expansion of its open AI ecosystem, releasing new models, datasets, and development tools to accelerate real-world AI across industries.
The company positioned the release as a push to move AI beyond digital tasks and into systems that can perceive, reason, and act in physical environments.
The announcement spans agentic AI, physical AI, robotics, autonomous vehicles, and biomedical research.
NVIDIA says the combined release gives developers one of the most comprehensive open AI toolkits currently available.
Reasoning-based autonomous driving
For autonomous vehicles, NVIDIA introduced Alpamayo, a new family of open models, simulation tools, and datasets designed for reasoning-based driving systems.
The release includes Alpamayo 1, an open vision language action model that enables vehicles to understand their surroundings and explain their actions.
NVIDIA also launched AlpaSim, an open-source simulation framework that supports closed-loop training and evaluation across varied environments and edge cases.
To support development, the company released Physical AI Open Datasets with more than 1,700 hours of driving data collected across a wide range of geographies and conditions.
Open models at scale
NVIDIA’s expanded portfolio includes open models from several product families.
These include Nemotron for agentic AI, Cosmos for physical AI, Alpamayo for autonomous vehicle development, Isaac GR00T for robotics, and Clara for healthcare and life sciences.
Alongside the models, NVIDIA released large-scale open datasets covering multiple domains.
These resources include 10 trillion language training tokens, 500,000 robotics trajectories, 455,000 protein structures, and 100 terabytes of vehicle sensor data.
The company says the scale and diversity of the data are intended to accelerate innovation across language systems, robotics, scientific research, and autonomous driving.
A growing list of companies has adopted NVIDIA’s open model technologies, including Bosch, Salesforce, ServiceNow, Palantir, Uber, Hitachi, CrowdStrike, Cohesity, Fortinet, and several robotics developers.
Physical AI and robotics
A central theme of the release focuses on physical AI, where systems must operate in complex real-world environments.
NVIDIA introduced new Cosmos open world foundation models designed to support perception, reasoning, and world generation.
Cosmos Reason 2 serves as a reasoning vision language model that improves how robots and AI agents interpret and interact with physical spaces.
NVIDIA also released Cosmos Predict 2.5 and Cosmos Transfer 2.5, which generate large volumes of synthetic video data across diverse conditions and environments.
The company paired these models with embodiment-specific tools. Isaac GR00T N1.6 targets humanoid robots and supports reasoning-driven, full-body control.
NVIDIA also released blueprints within its Metropolis platform to help developers build video search and summarization systems for large-scale video analysis.
Healthcare and deployment
NVIDIA also expanded its Clara platform with new AI models for biomedical research and drug discovery.
These tools support protein design, drug synthesis planning, safety testing, and RNA structure prediction.
A new dataset of 455,000 synthetic protein structures accompanies the release.
Developers can access NVIDIA’s open models and datasets through GitHub, Hugging Face, and build.nvidia.com.
Many models are also available as NVIDIA NIM microservices for deployment across cloud, edge, and on-premise systems.
We’re at the CES 2026!
You can explore all CES 2026 stories and coverage from the IE team by clicking here.
The Blueprint
