Nvidia has open sourced a new portfolio of AI models, tools and robotics hardware at CES, aiming to accelerate safe, transparent and scalable autonomy across vehicles and robots.
Nvidia Corp. has released more than half a dozen artificial intelligence models under open-source licences, targeting autonomous systems such as self-driving vehicles, humanoid robots and industrial robots. Announced at CES in Las Vegas, the release spans open source AI models, simulation and orchestration tools, and new robotics hardware, positioning Nvidia as a platform enabler rather than a closed-stack provider.
The flagship release is Alpamayo 1, a 10-billion-parameter Vision-Language-Action (VLA) model designed to train and evaluate autonomous driving systems. The model uses camera footage to generate driving trajectories and incorporates chain-of-thought reasoning, allowing it to break navigation tasks into smaller steps, explain its decisions and handle rare, complex driving scenarios. Nvidia emphasised that Alpamayo 1 is intended for training and validation, not direct deployment in vehicles, with larger models planned in the future.
“Alpamayo brings reasoning to autonomous vehicles, allowing them to think through rare scenarios, drive safely in complex environments and explain their driving decisions — it’s the foundation for safe, scalable autonomy,” said Jensen Huang, Chief Executive Officer, Nvidia.
Nvidia also expanded its Cosmos open source world foundation models with Cosmos Transfer 2.5, Cosmos Predict 2.5 and Cosmos Reason 2.0, supporting synthetic data generation, future object simulation and autonomous action execution. Cosmos Reason powers Isaac GR00T N1.6, a VLA model optimised for humanoid robots.
“Salesforce, Milestone, Hitachi, Uber, VAST Data and Encord are using Cosmos Reason for traffic and workplace productivity AI agents,” wrote Kari Briski, Vice President of Generative AI Software, Nvidia. “Franka Robotics, Humanoid and NEURA Robotics are using Isaac GR00T to simulate, train and validate new behaviors for robots before scaling to production.”
The open source stack is complemented by simulation tools AlpaSim and Isaac Lab-Arena, the OSMO workflow orchestrator, and the new Jetson T4000 Blackwell-based robotics computing module, reinforcing Nvidia’s push to accelerate open, transparent and scalable autonomy.
