Hugging Face's logo
Back to Articles

Building Autonomous Vehicles That Reason with the NVIDIA Alpamayo Open Ecosystem

Community Article Published January 5, 2026
Upvote
9
Marco Pavone's avatar
Marco Pavone
drmapavone

NVIDIA today released Alpamayo, an open ecosystem of models, simulation tools, and datasets to enable development of reasoning-based autonomous vehicle (AV) architectures. Our goal is to provide researchers and developers with a flexible, fast, and scalable platform for evaluating, and ultimately training, modern reasoning-based AV architectures in realistic closed-loop settings.

AV research is undergoing a rapid shift. The field is being reshaped by the emergence of reasoning-based vision–language–action (VLA) models that bring human-like thinking to AV decision-making. These models can be viewed as implicit world models operating in a semantic space, allowing AVs to solve complex problems step-by-step and to generate reasoning traces that mirror human thought processes. This shift extends beyond the models themselves: traditional open-loop evaluation is no longer sufficient to rigorously assess such models, and new evaluation tools are required.

For our first Alpamayo release, to help the community get started, we are providing the three key components needed to build reasoning AVs: a base model, a large-scale training dataset, and a simulation framework for testing and evaluation.

NVIDIA Alpamayo 1: An Open Reasoning VLA for AVs

Alpamayo 1 is an open, 10B-parameter chain-of-thought reasoning VLA model built on the Cosmos-Reason VLM backbone and designed for the autonomous vehicle research community. It takes multi-camera video as input and outputs both driving trajectories and reasoning traces that expose the logic behind each decision.

✨ Key Highlights

  • Local-friendly 10B-parameter size for rapid development and research.

  • State-of-the-art performance in multiple aspects (reasoning quality, trajectory accuracy, alignment, safety, latency, and more), with particular improvements observed due to the addition of reasoning.

  • Easy-to-use scripts and notebooks that enable immediate application across a wide range of use cases, from distillation into an online model to offline data curation.

🤖 Popular Use Cases

  • AV model distillation – Leverage the pretrained weights as an offline teacher to develop onboard-ready models (e.g., via output or feature supervision during training).

  • Data labeling and curation - Identify interesting scenarios and label them with plausible future trajectories and reasoning traces.

  • AV planning and reasoning – Generate trajectories and reasoning traces to evaluate the outputs of smaller, edge-deployed models or to explore alternative outcomes through multi-trajectory simulation rollouts.

NVIDIA Physical AI AV Dataset: Large-Scale, Diverse AV Data

The PhysicalAI-Autonomous-Vehicles dataset provides one of the largest, most geographically diverse collections of multi-sensor data for AV researchers to build the next generation of reasoning-based end-to-end driving systems. The associated physical_ai_av developer kit makes it easy to get started immediately with the dataset, containing a data interface and a detailed wiki about the format and structure of the dataset.

physicalaiav

✨ Key Highlights

  • Composed of 1,727 hours of driving (300,000+ clips, each 20 seconds long) recorded in 25 countries and 2,500+ cities (coverage shown below, color indicates the number of clips per country).

  • The dataset captures diverse traffic, weather conditions, obstacles, and pedestrian behaviors in their environment.

  • Sensor data includes multi-camera and LiDAR coverage for all clips, and radar coverage for more than half clips.

Picture_1

🤖 Popular Use Cases

  • AV model training and evaluation – Train and evaluate your end-to-end AV model on data from a variety of geographies and driving conditions.

  • Neural reconstruction and rendering – Leverage multi-sensor data to create and evaluate new neural reconstruction and rendering techniques.

  • Anomaly detection and safety evaluation – Explore how your model handles unusual events by evaluating it on unseen geographies, agent types, and behaviors.

NVIDIA AlpaSim: Closed-Loop Simulation for AV Evaluation

AlpaSim is an open-source end-to-end AV simulation platform designed specifically for research and development. It enables users to test end-to-end AV policies—particularly reasoning-based ones—in a closed-loop setting by simulating realistic sensor data, vehicle dynamics, and traffic scenarios within a modular and extensible testbed. AlpaSim is ready to use immediately, with reference implementations for all core services as well as 900+ reconstructed scenarios available on Hugging Face.

✨ Key Highlights

  • Clear, modular APIs via gRPC, making it easy to integrate new services without dependency conflicts.

  • Arbitrary horizontal scaling, allowing researchers to allocate compute where it matters most.

  • Strong GPU utilization and throughput due to the pipelined execution of multiple rollouts in parallel.

🤖 Popular Use Cases

  • Algorithm validation – Test new autonomous driving algorithms in realistic environments.

  • Safety analysis – Evaluate vehicle behavior in edge cases and challenging scenarios.

  • Performance benchmarking and regression testing – Compare different models and configurations.

  • Debugging – Understand and debug complex autonomous driving behaviors.

Conclusion

Reasoning models in AVs will unlock new capabilities and levels of safety for the next generation of autonomous systems. We hope that this introduction, and future Alpamayo releases, will accelerate the whole industry with new resources and tools.

Resources

For more information, please check out our technical blog at https://developer.nvidia.com/blog/building-autonomous-vehicles-that-reason-with-nvidia-alpamayo as well as the below resources.

Alpamayo 1:

Physical AI AV Dataset:

AlpaSim:

Research in Reasoning VLA Models, End-To-End AV Simulation and Training, and Physical AI Safety:

Community

Sign up or log in to comment

Upvote
9