DeepRoute.ai has announced the launch of DeepRoute IO 2.0, its latest smart driving platform powered by the company's advanced VLA (Vision-Language-Action) model. This next-generation platform marks a significant step toward delivering smarter, safer, and more human-like driving experiences for everyday users, the company claims.
DeepRoute IO 2.0 features a flexible multi-chip, multi-sensor design, supporting both LiDAR-equipped and pure vision configurations, enabling seamless integration across different vehicle models and automakers. It will first debut on NVIDIA DRIVE AGX Thor, running on DriveOS. DRIVE Thor is based on the Blackwell GPU architecture, purpose built for reasoning VLA models. DeepRoute.ai has already secured five confirmed OEM partnerships for the deployment of DeepRoute IO 2.0, with the first production vehicles scheduled to reach the market later this year.
According to Maxwell Zhou, CEO of DeepRoute.ai, "DeepRoute IO 2.0 adopts the VLA model integrated with the large language model, which brings two core advantages: Chain-of-thought reasoning and extensive knowledge base."
It is claimed the Chain-of-thought enables the system to analyze complex traffic scenarios through logical and human-like reasoning, enhancing the transparency and reliability of driving decisions. The extensive knowledge base allows the model to draw upon accumulated driving experience across diverse conditions, adapting seamlessly to real-world situations. Together, these capabilities enable DeepRoute IO 2.0 delivering reliable, safe, and highly adaptive assisted driving.
DeepRoute IO 2.0 platform, powered by the VLA model, significantly enhances the ability to manage complex driving scenarios. The platform delivers defensive driving by combining spatial understanding with advanced reasoning to handle diverse conditions and eliminate blind-spot risks. It ensures transparency through step-by-step explanations of its decision-making process, reducing the "black box" effect. With its OCR capability, the system can recognize traffic signs and interpret text road information in real time, just like a human driver. Finally, the voice control feature supports seamless interaction with humans using natural language, allowing drivers to control the vehicle through verbal commands.
DeepRoute.ai says it has validated DeepRoute IO 2.0 and the VLA model in real-world urban environments to ensure reliable performance under diverse conditions. The company continues to collaborate with global OEM partners to accelerate the deployment of smart driving technology in mass-production vehicles worldwide.
