Banma interview describes a move from cockpit assistants to vehicle agents

THE ESSENTIALS

Banma links AutoOmni and its Yan AI model family with agents that can use vehicle and external services.
The company says its broader software has reached more than 10 million vehicles; that is not a deployment count for every new agent.
The interview provides a strategy account rather than independent testing or a vehicle-by-vehicle release list.

A company interview outlines the vehicle-agent roadmap

Banma Intelligence describes a move from voice assistants toward software that can organize and execute tasks in cars, in an August 19 interview with TechNode. Its proposed stack combines Yan AI models, the on-device AutoOmni multimodal model and agent systems including SystemAgent, SuperAgent and AutoClaw.

The company gives parking payments and reminders about belongings left in a car as examples. These require perception, access to vehicle information and connections to services, rather than a language model alone. They are intended capabilities described in an interview, not features confirmed for every car running Banma software.

Existing reach and new features are different measures

Banma says it has worked with 69 automakers and that its solutions have reached more than 10 million vehicles. The company also describes adaptation to around 10 chip companies and more than 30 chip platforms. Those figures cover its broader business and do not establish the installed base of AutoClaw or AutoOmni.

The interview describes local processing for responsiveness, privacy and offline use, with cloud resources available for more demanding tasks. A completed parking payment still depends on the relevant service being connected and authorized. An offline model cannot by itself complete an online transaction.

What overseas readers can establish

The account names relationships with international and Chinese automakers, but provides no complete list of vehicle models, markets, software versions or enabled agent functions. It therefore establishes Banma's product direction, while leaving actual availability to each deployment.

For a buyer, the distinction is practical: a manufacturer's relationship with Banma is not a guarantee that a named car can carry out the interview's examples. The relevant release information is the supported task and version in the market where that vehicle will be used.