Nvidia Releases Alpamayo 2 Edge for Developers
The two-billion-parameter driving AI model that plans vehicle paths and explains its decisions is now available on Hugging Face.
Updated on Nov. 11, 2026 in Artificial Intelligence

Nvidia made Alpamayo 2 Edge available to developers on Hugging Face after announcing it on October 21 at GTC Berlin. The in-vehicle model takes in camera feeds, navigation data and vehicle movement to produce a planned trajectory and a reasoning trace.
Why it matters
The model is part of Nvidia's toolkit for Level 4 autonomous vehicles and robotaxis, giving developers a compact option for onboard driving systems. Its reasoning trace is designed to make the model's path decisions easier to inspect.
Nvidia's figures for Alpamayo 2 Edge are self-reported specifications: two billion parameters, roughly 10 Hz processing on DRIVE AGX Thor, and 160 to 190 milliseconds from camera image to planned path with TensorRT optimization. Independent performance validation on public roads has not been reported.
The players
Nvidia
Nvidia is a American semiconductor and AI computing company best known for its GPUs and its DRIVE autonomous-vehicle platform.
The details
Alpamayo 2 Edge takes in camera feeds, navigation data and the vehicle's recent movement, then outputs a planned trajectory along with an adaptive reasoning trace that explains the plan. It was developed from the larger Alpamayo 2 Super model through distillation.
Timeline
October 21, 2026: Nvidia announces Alpamayo 2 Edge at GTC Berlin.
November 11, 2026: Model made available to developers on Hugging Face.
Early 2027: Nvidia's Alpamayo Agentic Factory blueprint is expected.
The Tech Race
Alpamayo 2 Edge is measured on NVIDIA DRIVE AGX Thor, so its 10 Hz throughput extends Nvidia's in-vehicle compute stack into a smaller reasoning model. The release positions Nvidia against rivals building onboard driving AI for Level 4 autonomy and robotaxis.
Developers can now download the model from Hugging Face and test whether a compact reasoning model fits their onboard driving systems. For drivers, the practical effect will depend on whether automakers and robotaxi operators adopt it in production vehicles.
The takeaway
A smaller model that still explains its driving decisions could make autonomous systems easier to audit. Developers evaluating onboard AI should weigh the reported latency figures against their own hardware and road-test results.
What happens next
Nvidia's Alpamayo Agentic Factory blueprint for integrating AI agents is expected in early 2027.
Further reading
Explore more Artificial Intelligence coverage to see how autonomous driving models are evolving across the industry.
Source note: This article includes information reported by TokenPost.



