Banma Intelligence is promoting an AI-native approach to vehicle software centered on a system it calls Yan AI. According to a TechNode report, Yan AI spans foundation models, on-device models, AI agents, and an AI-native operating system. The company links specific product names to this structure, describing how the pieces are intended to align within its automotive software strategy. TechNode reports that the scope and organization of Yan AI are defined by these four areas, which Banma positions as the basis for how its software is designed to operate across cars and supporting infrastructure. TechNode reported the company’s strategy in detail.
Within that outline, Banma identifies several products that it associates with distinct roles. The company describes AutoOmni as an on-device multimodal model, and it names SystemAgent, SuperAgent, AutoClaw, and AIOS as parts of its product strategy. These descriptions are company statements. Together, the named elements are presented as components that Banma places under the Yan AI umbrella, connecting model capability, agent behavior, and operating system functions to what it calls an AI-native direction for vehicle software.
Yan AI architecture and components
TechNode reports that Yan AI is structured around four main pillars that Banma highlights. Foundation models are positioned as core capabilities. On-device models are described as adapted for in-vehicle execution. AI agents are identified as part of how intelligence is organized for tasks and interactions. An AI-native operating system is included as the system layer meant to support and coordinate how these elements work together. Banma presents this arrangement as a way to link model development and runtime operations across the vehicle and the cloud.
In this organization, AutoOmni is described by the company as an on-device multimodal model that fits into the vehicle-side portion of its approach. Banma also names SystemAgent and SuperAgent among the AI agent elements it cites, and it includes AutoClaw and AIOS as parts of the broader product strategy. These labels are offered by the company to indicate where certain capabilities are intended to reside within the Yan AI structure. The descriptions of these product capabilities are company statements.
Banma’s naming of agents, models, and operating system elements is presented to convey a consistent stack. The company links the categories to roles that are meant to connect data and compute on the car with external resources when needed. By identifying a foundation model layer, an on-device model layer, AI agents, and an AI-native operating system, Banma sets out how it describes the flow from core model capacity to software behavior in vehicles. The emphasis throughout remains on the company’s definitions of each part and how those parts are grouped under Yan AI.
Banma’s strategy sits alongside other Chinese automotive AI efforts, including Pony.ai’s NVIDIA DRIVE Hyperion platform.
Cloud-device collaboration model
Banma argues that vehicles need a collaboration model that assigns work between the car and the cloud. The company’s position is that real-time and privacy-sensitive tasks should run on the vehicle, while more complex tasks should run in the cloud. This division is described as central to how its AI-native approach is intended to function. In Banma’s view, the car is the place for immediate computation that benefits from low latency and local handling of sensitive data, and the cloud is where heavier processing can be carried out.
This argument is presented as part of the same overall strategy that links Yan AI’s components to the way computation is allocated. Banma ties the on-device portion of its system to time-critical and privacy-focused tasks, and it points to cloud resources for tasks it considers better suited to external compute. The company aligns its named elements with this split to indicate how the components should interact across vehicle hardware and remote infrastructure. These statements describe how Banma says its system is designed to coordinate resources.
For readers seeking broader context on software and compute in mobility, see a previous EastFrontier analysis. The reference provides a look at how different actors in the space discuss the relationship between onboard systems and external computing. Banma’s argument focuses on a car-first stance for immediate operations, with cloud support reserved for tasks it classifies as more complex. The collaboration model described by the company is intended to connect its model layers, agent behavior, and AI-native operating system to a clear division of processing responsibilities.
Scale and ecosystem statements
Banma says it has worked with 69 automakers. The company also says it has deployed solutions in more than 10 million intelligent vehicles. In addition, Banma says it has adapted its technology to about 10 chip companies and over 30 chip platforms. These figures are company statements. The numbers are presented by Banma as part of how it describes the reach of its approach across vehicle makers and computing platforms.
The company’s scale statements position its software as present across a range of partners and hardware environments. By citing work with 69 automakers and deployment in more than 10 million intelligent vehicles, Banma sets out a claim about the breadth of its solutions. By stating that its technology has been adapted to about 10 chip companies and over 30 chip platforms, it highlights the variety of silicon contexts it says it supports. These statements are provided by the company and are presented as such.
These scale claims are linked by Banma to the Yan AI strategy and the product names it lists. The company describes AutoOmni, SystemAgent, SuperAgent, AutoClaw, and AIOS as parts of its product strategy, and it connects the collaboration model between car and cloud to how it says these elements are intended to work. The figures supplied by Banma serve to underscore the scope it associates with that strategy. The product capabilities and scale figures referenced here are company statements.
Taken together, the elements described by Banma outline its AI-native vision for automotive software. Yan AI is presented as a system spanning foundation models, on-device models, AI agents, and an AI-native operating system. The company argues for a collaboration model that keeps real-time and privacy-sensitive computation on the vehicle and places more complex tasks in the cloud. Alongside those points, Banma cites scale claims across automakers and chip platforms. All of these descriptions are company statements that define how Banma says it is building and positioning its automotive software.