PATEO Deploys Doubao and Nemotron in Thor-Based Vehicle AI Box

PATEO announced on November 2, 2026 that it is deploying NVIDIA Nemotron and ByteDance’s 35-billion-parameter Doubao language model in an AI Box built on NVIDIA DRIVE AGX Thor, combining local AI inference with an agent platform for vehicle cockpits.

PATEO has also secured a project nomination from an unnamed leading new-energy vehicle automaker for a new-generation vehicle. It expects mass production for that program in the second quarter of 2027, giving the AI Box a defined vehicle program while production remains a future milestone. At Auto China 2026, PATEO announced joint development of the solution with ByteDance and an unnamed leading new-energy vehicle automaker.

Local Inference and Cabin Agents

The in-vehicle suite uses NVIDIA Nemotron 3.5 ASR for speech recognition and, through collaboration with ByteDance, Doubao’s mixture-of-experts language model. Text-to-speech synthesis and a separate 2-billion-parameter language model for semantic classification and request rejection round out the suite.

PATEO said the integrated suite can run inference locally without a cloud connection, supporting multilingual voice interaction and real-time streaming speech recognition. That design keeps core AI functions available when connectivity is limited, while connected services and cloud updates extend the system’s capabilities.

The AI Agent platform combines voice, visual and sensor inputs to interpret user requests and cabin conditions. Within vehicle safety permissions, it can invoke cabin controls and online services, coordinate concurrent tasks and manage longer-running requests. Preference and habit memory for personalized interactions requires user authorization.

Proactive Features and Hardware Design

PATEO’s examples show the agent moving beyond question-and-answer voice control. Sentinel AI can identify and issue alerts for parking violations, flooding and fire. Driver and passenger climate settings can be adjusted separately through voice commands or gestures, while fatigue detection can prompt suggestions to open a window or activate seat massage during extended highway driving. In areas with poor air quality, the system can suggest closing windows and switching to recirculation based on location and learned driving habits.

The hardware rationale centers on what happens as those interactions accumulate. PATEO describes the AI Box as connecting perception, reasoning, execution and learning, with memory capacity and bandwidth mattering alongside peak compute measured in TOPS. As conversation turns and agent actions add context during a trip, the model’s key-value cache grows and increases demands on memory. PATEO said Thor’s accelerated compute and high memory bandwidth help support responsive local inference.

Modular software and integration with vehicle systems and NVIDIA’s edge-and-cloud AI ecosystem support model deployment, tuning and customization across vehicle programs. The AI Box can also add dedicated AI computing without replacing a vehicle’s original main control chip. That offers an upgrade path for existing vehicles whose processors cannot support larger models, concurrent multimodal interactions and proactive agents through software updates alone.

A Services Strategy Beyond the AI Box

The business plan extends beyond selling the hardware. PATEO’s in-vehicle Tokenomics strategy aims to generate recurring revenue around the deployment and use of AI capabilities, adding ongoing services and operations to the initial hardware transaction.

In June 2026, PATEO partnered with Mininglamp Technology, Xunce Technology and Saimo Technology to develop a token-based physical-AI world model and TokenOS. That work combines on-device hardware, simulation testing and data infrastructure, creating a foundation for delivering and operating AI services.

PATEO is also applying the platform outside cars. It reported a project nomination from an unnamed leading domestic robotics computing provider, creating an opportunity to reuse its intelligent-interaction and hardware-engineering capabilities in physical-AI applications.

For the AI Box, the central commercial test is to move the nominated vehicle program into production while showing that a reusable technical platform can also support the recurring services revenue envisioned by Tokenomics.