SAN JOSE, Calif., March 17, 2026 /PRNewswire/ -- At NVIDIA GTC 2026, DeepRoute.ai presented a comprehensive introduction to its 40-billion-parameter Vision-Language-Action (VLA) Foundation Model architecture, representing a fundamental breakthrough in autonomous driving development. The model introduces a unified architecture that integrates perception, reasoning, and action, enabling systems not only to drive, but to understand and evaluate their own decision-making in real time.
DeepRoute.ai has already achieved significant commercial success, having delivered its advanced autonomous driving systems across more than 250,000 production vehicles. In October 2025, DeepRoute.ai captured nearly 40% market share among third-party suppliers in the high-level autonomous driving segment for a single month. Building on this momentum and fueled by the continuous evolution of its Foundation Model, the company is targeting deployment of one million vehicles equipped with its advanced driving solutions by the end of 2026.
Breaking the Bottleneck: From Days to Hours
Autonomous driving development has long been hampered by the inefficiencies of traditional "data closed-loop" workflows. In conventional systems, data must be manually collected, reviewed, annotated, and retrained—a process that typically requires more than five days per iteration. Meanwhile, companies accumulate vast volumes of raw driving data, most consisting of routine scenarios that offer limited training value and can even degrade model performance.
"At its core, autonomous driving is a scaling problem," said Tongyi Cao, CTO of DeepRoute.ai. "While the industry has made significant progress, true large-scale deployment remains elusive because traditional execution paths are flawed. The bottleneck is no longer about acquiring data; it is about how efficiently a system can filter out the noise and convert massive amounts of raw data into high-value training samples."
DeepRoute.ai's solution: compress the data processing cycle from over five days to approximately 12 hours through intelligent automation.
One Model, Three Roles: Driver, Analyst, and Critic
The 40B VLA Foundation Model performs three complementary functions simultaneously:
The Driver – Executes real-time driving actions based on visual inputs
The Analyst – Identifies critical driving events and explains decisions through causal reasoning
The Critic – Evaluates trajectories for safety, comfort, and human-like behavior
"Our solution to the industry's scaling bottleneck is a unified, 40-billion parameter Vision-Language-Action Foundation Model," Cao explained. "This model goes beyond basic vehicle control. It possesses the capability to analyze data and evaluate driving behavior. Simply put, this model serves not only as the 'driver,' but simultaneously as the 'analyst' and the 'critic.'"
By embedding these capabilities within a single foundation model, DeepRoute.ai has automated large portions of the data pipeline. The system autonomously identifies high-value events such as near-misses and rare scenarios, performs root-cause analysis, and generates reasoning annotations, all without manual intervention.
A Self-Evolving Data Flywheel
The architecture enables a self-reinforcing development cycle where improvements in driving performance directly enhance the system's ability to process and curate its own training data.
"Traditional data closed loops are highly dependent on manual human processes, which severely limits iteration speed," said Cao. "By leveraging our Foundation Model, we have entirely reconstructed this workflow. The model autonomously handles data mining, reason diagnosis, and behavior scoring. Every single iteration of this workflow compounds directly into a measurable enhancement of our AI capabilities."
This self-evolving flywheel accelerates capability growth while dramatically reducing reliance on manual labeling.
Scale and Momentum: 250K to 1M Vehicles
"By the end of 2025, we successfully delivered over 250,000 mass-produced vehicles equipped with DeepRoute.ai's autonomous driving systems," Cao said. "The Foundation Model serves as the core cornerstone for DeepRoute.ai's next-generation autonomous driving assistance and functions as a fundamental AI framework for the physical world. This unified architecture enables the system to go beyond mere execution; it understands complex traffic environments, explains the underlying logic of its decisions, and evaluates driving behaviors. This evolution provides its autonomous driving systems with more comprehensive cognitive and decision-making capabilities."
Through its presentation at GTC 2026, DeepRoute.ai demonstrated how its 40B Vision-Language-Action Foundation Model architecture is accelerating the path to scalable, safe autonomous driving through continuous, data-driven learning and rapid iteration cycles.
About DeepRoute.ai
DeepRoute.ai is a leading artificial intelligence company developing advanced autonomous driving systems. Driven by the vision of achieving Artificial General Intelligence (AGI) for the physical world, the company leverages state-of-the-art foundation models to deliver highly reliable, safety-first autonomous driving solutions. Backed by top-tier investors with over $700 million in funding, DeepRoute.ai has successfully deployed its systems across more than 200,000 mass-produced consumer vehicles. By prioritizing scalable and innovative smart mobility, the company is establishing a robust foundation to pioneer the future of commercial Robotaxi operations.
SAN JOSE, Calif., March 17, 2026 /PRNewswire/ -- At NVIDIA GTC 2026, DeepRoute.ai presented a comprehensive introduction to its 40-billion-parameter Vision-Language-Action (VLA) Foundation Model architecture, representing a fundamental breakthrough in autonomous driving development. The model introduces a unified architecture that integrates perception, reasoning, and action, enabling systems not only to drive, but to understand and evaluate their own decision-making in real time.
DeepRoute.ai Presents 40B Vision-Language-Action Foundation Model at NVIDIA GTC 2026, Accelerating Autonomous Driving at Scale
DeepRoute.ai Presents 40B Vision-Language-Action Foundation Model at NVIDIA GTC 2026, Accelerating Autonomous Driving at Scale
DeepRoute.ai Presents 40B Vision-Language-Action Foundation Model at NVIDIA GTC 2026, Accelerating Autonomous Driving at Scale
DeepRoute.ai Presents 40B Vision-Language-Action Foundation Model at NVIDIA GTC 2026, Accelerating Autonomous Driving at Scale
DeepRoute.ai has already achieved significant commercial success, having delivered its advanced autonomous driving systems across more than 250,000 production vehicles. In October 2025, DeepRoute.ai captured nearly 40% market share among third-party suppliers in the high-level autonomous driving segment for a single month. Building on this momentum and fueled by the continuous evolution of its Foundation Model, the company is targeting deployment of one million vehicles equipped with its advanced driving solutions by the end of 2026.
Breaking the Bottleneck: From Days to Hours
Autonomous driving development has long been hampered by the inefficiencies of traditional "data closed-loop" workflows. In conventional systems, data must be manually collected, reviewed, annotated, and retrained—a process that typically requires more than five days per iteration. Meanwhile, companies accumulate vast volumes of raw driving data, most consisting of routine scenarios that offer limited training value and can even degrade model performance.
"At its core, autonomous driving is a scaling problem," said Tongyi Cao, CTO of DeepRoute.ai. "While the industry has made significant progress, true large-scale deployment remains elusive because traditional execution paths are flawed. The bottleneck is no longer about acquiring data; it is about how efficiently a system can filter out the noise and convert massive amounts of raw data into high-value training samples."
DeepRoute.ai's solution: compress the data processing cycle from over five days to approximately 12 hours through intelligent automation.
One Model, Three Roles: Driver, Analyst, and Critic
The 40B VLA Foundation Model performs three complementary functions simultaneously:
The Driver – Executes real-time driving actions based on visual inputs
The Analyst – Identifies critical driving events and explains decisions through causal reasoning
The Critic – Evaluates trajectories for safety, comfort, and human-like behavior
"Our solution to the industry's scaling bottleneck is a unified, 40-billion parameter Vision-Language-Action Foundation Model," Cao explained. "This model goes beyond basic vehicle control. It possesses the capability to analyze data and evaluate driving behavior. Simply put, this model serves not only as the 'driver,' but simultaneously as the 'analyst' and the 'critic.'"
By embedding these capabilities within a single foundation model, DeepRoute.ai has automated large portions of the data pipeline. The system autonomously identifies high-value events such as near-misses and rare scenarios, performs root-cause analysis, and generates reasoning annotations, all without manual intervention.
A Self-Evolving Data Flywheel
The architecture enables a self-reinforcing development cycle where improvements in driving performance directly enhance the system's ability to process and curate its own training data.
"Traditional data closed loops are highly dependent on manual human processes, which severely limits iteration speed," said Cao. "By leveraging our Foundation Model, we have entirely reconstructed this workflow. The model autonomously handles data mining, reason diagnosis, and behavior scoring. Every single iteration of this workflow compounds directly into a measurable enhancement of our AI capabilities."
This self-evolving flywheel accelerates capability growth while dramatically reducing reliance on manual labeling.
Scale and Momentum: 250K to 1M Vehicles
"By the end of 2025, we successfully delivered over 250,000 mass-produced vehicles equipped with DeepRoute.ai's autonomous driving systems," Cao said. "The Foundation Model serves as the core cornerstone for DeepRoute.ai's next-generation autonomous driving assistance and functions as a fundamental AI framework for the physical world. This unified architecture enables the system to go beyond mere execution; it understands complex traffic environments, explains the underlying logic of its decisions, and evaluates driving behaviors. This evolution provides its autonomous driving systems with more comprehensive cognitive and decision-making capabilities."
Through its presentation at GTC 2026, DeepRoute.ai demonstrated how its 40B Vision-Language-Action Foundation Model architecture is accelerating the path to scalable, safe autonomous driving through continuous, data-driven learning and rapid iteration cycles.
About DeepRoute.ai
DeepRoute.ai is a leading artificial intelligence company developing advanced autonomous driving systems. Driven by the vision of achieving Artificial General Intelligence (AGI) for the physical world, the company leverages state-of-the-art foundation models to deliver highly reliable, safety-first autonomous driving solutions. Backed by top-tier investors with over $700 million in funding, DeepRoute.ai has successfully deployed its systems across more than 200,000 mass-produced consumer vehicles. By prioritizing scalable and innovative smart mobility, the company is establishing a robust foundation to pioneer the future of commercial Robotaxi operations.
** This press release is distributed by PR Newswire through automated distribution system, for which the client assumes full responsibility. **
HONG KONG, March 17, 2026 /PRNewswire/ -- As outdoor living spaces continue to gain popularity among homeowners, demand for reliable and weather-resistant lighting has increased significantly. In response to this trend, Dazuma has expanded IP65 protection across a broader range of its outdoor lighting collection, reinforcing the brand's long-standing focus on durability, safety, and consistent product quality.
Dazuma has observed a clear rise in demand for waterproof outdoor lighting, reflected in increased customer inquiries about IP ratings, closer comparisons of technical specifications during the buying process, and growing interest in fixtures designed for long-term outdoor use. According to the company, buyers today are more likely to review technical specifications before making a purchase, with IP65 protection becoming one of the most frequently requested features for outdoor lighting.
IP65 is widely recognized as a high level of protection, meaning a fixture is fully protected against dust intrusion and resistant to water jets from multiple directions, making it suitable for everyday outdoor use. Because lighting installed on exterior walls, walkways, and garden areas must withstand rain, dust, and humidity, more homeowners now pay close attention to waterproof ratings, often using IP codes as a key reference when comparing outdoor fixtures.
"As outdoor living continues to grow in popularity, customers are paying closer attention to durability and waterproof performance when selecting outdoor lighting," said a spokesperson for Dazuma. "Expanding IP65 protection across more of our outdoor lighting collection helps ensure our products can perform reliably in real-world outdoor environments."
In response to evolving customer expectations, Dazuma has expanded IP65 protection across a wider range of its outdoor lighting products, including:
This approach helps maintain consistent protection standards across different product categories while ensuring dependable performance in a variety of outdoor environments.
To support this expansion, Dazuma has refined its sourcing strategy and strengthened coordination with supply partners to ensure more outdoor lighting designs meet consistent protection requirements. The company has introduced clearer waterproof performance guidelines for exterior fixtures, requiring suppliers to provide detailed technical specifications and IP test documentation based on recognized protection standards. Products intended for outdoor use are also reviewed more carefully during the selection process, with IP65 ratings becoming an important reference when evaluating durability and suitability for real-world environments.
At the same time, Dazuma has worked more closely with its supply partners to align product requirements across different categories. By maintaining regular communication on technical specifications, confirming protection levels during product development, and applying more unified evaluation criteria, the company aims to ensure that outdoor wall lights, pathway lights, garden fixtures, and other exterior lighting designs meet consistent IP65 expectations while maintaining the overall style and variety of the collection.
The company noted that this move reflects its long-term commitment to product quality rather than a short-term adjustment. Over the years, Dazuma has continued refining its outdoor lighting designs to improve resistance to weather, corrosion, and environmental wear. Expanding IP65 protection across a broader range of fixtures is seen as a natural step in maintaining consistent quality across the brand's growing outdoor lighting lineup.
Looking ahead, Dazuma plans to gradually make IP65 protection a standard feature across its outdoor lighting collection, ensuring that fixtures designed for exterior use meet reliable performance expectations by default. As customer requirements continue to evolve, the company also expects to explore higher protection ratings for products intended for more demanding outdoor environments.
About Dazuma
Founded in France in 2016, Dazuma is a lighting and home décor brand focused on providing modern, functional, and well-designed products for indoor and outdoor living spaces.
Supported by a global supply chain, the company offers a wide range of lighting solutions, including indoor and outdoor lighting, wall lights, ceiling lights, ceiling fans, and decorative lamps, serving both homeowners and business customers across the United States. Dazuma is committed to delivering reliable quality at competitive prices while also providing customized lighting solutions for large-scale projects and B2B customers.
For more information, visit dazuma.us or follow Dazuma on social media:
