Author | Zhang Lianyi Editor | Jiang Tianshuang
As mobile physical AI moves toward cross-vertical mass production, how much commercial value can a unified "brain" release?
On October 29, Zhuoyu Technology officially launched its intelligent driving assistance solution for commercial vehicles, Zhuoyu Panxing™, under the theme "New Journey."
The launch event disclosed Panxing's deployment progress across multiple categories, including heavy-duty trucks, buses, light trucks, vans, and unmanned logistics vehicles. It also announced comprehensive cooperation with China's top six commercial heavy truck brands (XCMG, Shaanxi Auto, Sinotruk, FAW Jiefang, Foton, and Dongfeng). These six brands collectively account for over 93% of the domestic commercial heavy truck market share.
From "Chengxing" to "Panxing," and from passenger vehicles to commercial vehicles, Zhuoyu is not only expanding its business boundaries but also attempting to answer a critical question: How much commercial and social value can a unified AI "brain" unlock?
After all, Shen Shaojie, CEO of Zhuoyu Technology, hopes that Panxing will not only improve user experience but also enable intelligent mobility to create value, making mobile physical AI a real productivity force. This AI productivity is not about replacing drivers, but about gently supporting truck drivers with AI.
01 Passenger Vehicles as the "Ammunition Depot"
Intelligent driving in commercial vehicles is difficult due to the dual pressure of "physical limits" and "cost red lines."
Facing these pain points, the industry typically adopts two approaches: one focuses on heavy trucks, developing a dedicated system for them; the other enters from other scenarios, attempting to cover multiple vehicle types with a single solution.
Zhuoyu clearly chose the second approach, implementing migration and adaptation.
Data-wise, Zhuoyu has indeed accumulated strong confidence from its passenger vehicle base.
At the launch event, Zhuoyu disclosed the latest progress in its passenger vehicle business: currently serving 22 passenger vehicle clients, with over 60 models in mass production and over 110 models designated for production. Recently, it secured partnerships with international giants like Toyota and Hyundai. Its "Chengxing" brand's high-comprehension large model has evolved to version 4.1, with plans to mass-produce a native multimodal foundation model by Q12027.
In addition, the passenger vehicle L3 architecture solution was showcased at the Beijing Auto Show, with substantial progress expected in the first half of 2027.
This constitutes the core asset of "Panxing."
However, successful experiences in passenger vehicles cannot be directly transferred to commercial vehicles. A fully loaded heavy truck weighs approximately 40 tons, travels at 90 km/h on highways, features a long body, numerous blind spots, and long braking distances. Significant shifts in center of gravity between empty and full loads, combined with high-intensity operations, pose genuine technical challenges.
Liu Sikang, Vice President of Zhuoyu's Commercial Vehicle Business, presented the methodology of "unified underlying layer, scenario enhancement, and engineering adaptation": migrating the underlying models from passenger vehicles, feeding them with data specific to heavy trucks, and solving heavy truck physical challenges through engineering means.
Specifically, at the model level, Zhuoyu migrated its passenger vehicle intelligent driving models to heavy truck scenarios, inheriting over 3 billion kilometers of cumulative data and nationwide mass-production generalization capabilities from passenger vehicles, enhanced by 400 million frames of heavy-truck-specific data.
In the future, Zhuoyu plans to extend these capabilities to scenarios frequently traveled by truckers, such as national and provincial roads, achieving more comprehensive coverage of the main road network.
"We let the same 'brain' learn to drive commercial heavy trucks," Liu Sikang stated. In his view, Zhuoyu broke the status quo where "end-to-end solutions could only be implemented in passenger vehicles," successfully achieving mass production of end-to-end large models for commercial heavy trucks.
Based on this, Zhuoyu released a series of products at the launch event:
Version System: PDS4 (High-Comprehension Large Model 4.0+) focuses on Highway NOA, meeting trunk line logistics needs, currently in mass production across multiple brands and reaching mature delivery status; PDS5 (Native Multimodal Foundation Model) targets buses, light trucks, and vans, expanding Urban NOA.
Sensing Hardware: First release of Jimu 2.0, installed inside the cabin to prevent rain and stone impacts, with a highway detection range of up to 500 meters, specifically addressing the pain points of heavy trucks "not seeing clearly and not stopping quickly."
Dynamics Control: Addressing unique characteristics such as "tail swinging" and inner wheel difference in 17-meter trailers, real-time visual recognition of trailer posture and estimation of the angle between tractor and trailer are used for precise dynamics tuning.
Releasing technical accumulation from the passenger vehicle domain into the heavy truck domain all at once.
02 Calculating the "Value Account" for Heavy Trucks with AI
Technology alone is not enough. Unlike passenger vehicle intelligent driving, which emphasizes "experience," commercial vehicle intelligent driving pays closer attention to the "economic account."
Zhuoyu Technology's core killer feature is its AI reinforcement learning fuel-saving algorithm. Fuel costs account for 30%-40% of operating costs for heavy trucks. Panxing uses simulation training to teach AI to find the "golden fuel cut-off point" like an experienced driver.
According to Zhuoyu's statistics, over the past two months, Zhuoyu tracked more than 200 mass-produced vehicles, accumulating over 20 million kilometers. With the AI fuel-saving algorithm equipped, vehicle fuel consumption continued to decrease, with measured fuel savings ranging from 4.5% to 7.5%. Based on current oil prices, this saves approximately CNY 10 per 100 km, CNY 80 per day for800 km driven, and an estimated CNY 20,000 to 30,000 per year.
Liu Sikang stated that CNY 20,000 to 30,000 per year represents profit for fleets and tangible income for drivers. Beyond fuel savings, reduced driver fatigue and improved labor efficiency, along with lower accident risks enhancing safety, may further reduce insurance premiums. Comprehensive calculations suggest that one set of intelligent driving assistance systems can pay for itself within a year.
This is the core driving force behind the commercial closed loop. However, it must be clarified that the actual effectiveness of the fuel-saving algorithm highly depends on driver habits, road conditions, and vehicle load. The extent to which measured data can be stably reproduced is key to winning B-end customers.
Beyond just the "economic account," Shen Shaojie mentioned stories of truck drivers from the book Rolling Wheels during the launch event. China has over 30 million commercial vehicle drivers who run a sophisticated "algorithm" in their brains but face long-term high-intensity driving and physical wear.
"A safe, comfortable, and cost-effective assisted driving system allows 30 million truckers to travel less burdened; we have no reason not to build it," Shen said.
This statement carries both humanistic value and commercial judgment: In the past year, China's freight volume reached 58.7 billion tons, with nearly three-quarters completed by commercial vehicle drivers on public roads.
In his view, China's commercial heavy truck market has entered an era of stock competition and structural replacement, making intelligent upgrades a natural differentiation opportunity. Meanwhile, commercial vehicle exports exceeded one million units in 2025, forming a full industrial chain layout of "complete vehicles + components + services" for Chinese commercial vehicles going global, presenting significant opportunities.
Of course, Shen Shaojie emphasized that this AI productivity is not about replacing drivers, but about gently supporting truck drivers with AI.
03 Extending Beyond Heavy Trucks to Full-Link L4
"Panxing" is an important part of Zhuoyu's "Intelligent Everything Mobile" strategy, but not the whole picture.
At the launch event, Zhuoyu clarified the "Four Completes" landscape for its commercial vehicle business: full product type coverage (heavy trucks, buses, light trucks, vans), full powertrain adaptation, full partnership with leading brands, and full domestic and overseas layout.
This is not just a slogan; it is already being implemented.
In terms of heavy trucks, the Shaanxi Auto Delong X5000 equipped with PDS4 has been mass-produced and is running on Yunda trunk lines, with Highway NOA usage expected to exceed 90% by year-end; XCMG Hanjing new energy heavy truck mass-production models are expected to be delivered by the end of this year; Sinotruk cooperative models are expected to roll off the line by year-end; Foton Auman Galaxy5 will hit the market by year-end; Dongfeng Tianlong KH and KL models have secured designations, with the Dongfeng Tianlong KH L2+ intelligent tractor equipped with PDS4 expected to enter mass production by the end of this year.
Regarding buses and light trucks, the Yutong Yuwei E9PRO City Edition will be equipped with PDS5, based on the NVIDIA Thor platform, supporting both Highway and Urban NOA; BAIC Foton's new light truck and van platform models will also be equipped with PDS5, targeting urban distribution scenarios.
In addition, the ROBOVAN jointly developed by Zhuoyu, SAIC Maxus, and Yisen Power has entered trial operation, while collaborations with other ecosystem partners are actively progressing.
In terms of globalization, Zhuoyu has reached overseas cooperation agreements with partners such as Sinotruk, Foton Motor, and Yutong Bus, covering full categories including commercial heavy trucks, buses, and light trucks.
This means Zhuoyu's intelligent map is expanding from "trunk line logistics" to "urban distribution" and even "last-mile delivery": from heavy trucks on highway trunk lines, to light trucks and vans in urban distribution, to ROBOVANS in last-mile delivery, Zhuoyu is building an intelligent mobile network covering all logistics scenarios.
The technological foundation supporting this extension is the Native Multimodal Foundation Model (MFM) that Zhuoyu is advancing. Enabling the same model to accurately understand the environment and adapt to the "driving style" of different carriers is the core challenge of cross-vertical deployment. The goal of MFM is to solve sensor multi-configuration and cross-carrier adaptation issues, truly realizing unified intelligent capabilities across verticals, regions, and scenarios, allowing one "AI brain" to serve all-scenario mobility.
Of course, the commercial vehicle market has extremely strong customization attributes. Adapting heavy trucks to multiple brands and models in a short time poses a great test of engineering delivery capabilities. Moving unmanned logistics vehicles from trial operation to scale involves crossing thresholds such as regulatory restrictions on open urban roads, complex last-mile delivery scenarios, and balancing unit cost with operational efficiency.
But the direction is clear: Zhuoyu is using a full-link layout of "trunk line + urban distribution + last mile" to connect commercial vehicle intelligent driving from a single-scenario "point" into a "network" covering the entire logistics chain.