Autobrains Brings NVIDIA Alpamayo Intelligence to the Road

Autobrains announced an expansion of its work with NVIDIA, incorporating the reasoning capabilities of NVIDIA Alpamayo 2 Super into real-world autonomous driving. Using its Agentic AI architecture, Autobrains distills Alpamayo’s foundation-model intelligence into specialized Driving Agents that reason and act in real time in production vehicles.

NVIDIA Alpamayo 2 Super is a 34B parameter foundation model that can reason through complex driving situations, generate trajectories and causal reasoning traces, and provide advanced training signals. Autobrains’ Agentic AI architecture turns this intelligence into real-world execution, activating only the agents relevant to each situation and enabling advanced reasoning in the vehicle without requiring the complete foundation model to run inside the car.

“Reasoning AI is the next great leap in autonomous driving,” said Rishi Dhall, vice president of automotive at NVIDIA. “Autobrains’ Agentic AI transforms advanced reasoning capabilities into specialized Driving Agents optimized for real-time execution, while NVIDIA Alpamayo provides the foundational intelligence to understand and reason through complex driving situations. Together, Autobrains’ Agentic AI and NVIDIA Alpamayo create a powerful path from frontier AI reasoning to intelligent, scalable autonomous driving in production vehicles.”

“For decades, autonomous driving has advanced by collecting more road data and building increasingly larger models,” said Igal Raichelgauz, founder and CEO of Autobrains. “Alpamayo and Autobrains’ Agentic AI change that equation. Foundation AI can now reason through rare and complex scenarios, while Agentic AI distills that intelligence into specialized Driving Agents that run efficiently in the vehicle. This shifts autonomous driving from a data flywheel to an AI flywheel, a fundamental inflection point for autonomous driving.”

The next breakthrough in autonomous driving is not only about building larger models. It is about turning their intelligence into systems that can operate within automotive compute, latency, power and cost constraints.

Autobrains distills Alpamayo into specialized Driving Agents. A context-aware orchestrator activates only the agents and skills required for each situation, preserving task-level quality while reducing compute requirements.

This creates a practical path from frontier-model reasoning to real-time execution without requiring the complete foundation model to run inside the vehicle.

Autonomous driving development has traditionally depended on collecting more road data, manually labeling it and retraining increasingly large models. Rare and complex scenarios remain the greatest challenge because they may take millions of road miles to encounter.

Alpamayo brings trajectories, causal reasoning traces and advanced training signals into Autobrains’ Learning by Thinking process. This knowledge helps generate challenging scenarios, build specialized Driving Agents and identify their weaknesses.

Each cycle allows AI to develop, challenge and improve other AI, accelerating progress across the situations that are hardest to capture on the road. Real-world data and validation remain essential, but they are no longer the only engines of improvement.

Autobrains and NVIDIA already collaborate on Level 4 programs with VinFast in Southeast Asia and Uber in Munich, both using NVIDIA Hyperion, an integrated compute and sensor platform powered by NVIDIA DRIVE AGX Thor, with NVIDIA Halos safety system to accelerate safe, production-ready autonomy.

Alpamayo expands the collaboration from the vehicle platform into the development of driving intelligence. Autobrains connects NVIDIA’s foundation-model reasoning with its Agentic AI architecture and real-world autonomous driving programs, creating a scalable path across passenger vehicles and robotaxis.

Autobrains is an AI company bringing autonomous driving to every car through its Agentic AI technology – a new approach to driving intelligence designed to operate on standard sensors and automotive-grade compute. The company holds over 300 patents and is working with leading OEMs and Tier 1 suppliers to enable scalable, cost-efficient ADAS and autonomous driving.

Photo: Autobrains