On July 27, during the 2025 World Artificial Intelligence Conference, Banma Intelligent Driving, together with Tongyi and Qualcomm, launched the first end-side multimodal large model solution, pushing the car's smart cockpit into the era of active intelligence.
"This is the industry's first end-side multimodal large-model solution based on the Qualcomm 8397 platform. It can achieve 90% of the 'perception-decision-execution' service closed loop of the smart cockpit through a purely vehicle-side approach." Banma Smart Driving Chief Technology Officer Si Luo said at the press conference that the end-native intelligent agent based on this solution can perform multimodal intent perception, understanding and interaction, realizing a generational upgrade from "command recipient" to "dialogue participant."
Robot delivering coffee
At the Banma Smart Driving booth, users could easily activate the local life agent in their Zhiji vehicles, equipped with Yuanshen AI. From ordering coffee to selecting specifications and flavors, and finally placing and paying, the entire process was conducted through natural conversation. A Yuanshen AI coffee shop robot delivered coffee to the car, giving users a glimpse into the symbiotic relationship between humans and machines.
Banma Smart Driving demonstrated its edge-side multimodal large model solution based on Qualcomm 8397 using scenario cases: when a user gets in the car sweating, Yuanshen AI will proactively turn on the air conditioner based on the cabin environment, user status and actions, bringing extremely cool or warm air; when encountering continuous congestion during travel, Yuanshen AI will proactively recommend a playlist after "seeing" it to alleviate congestion anxiety.
Qualcomm 8397 terminal model stand
From the perspective of the domestic automobile market, "big models on cars" are gradually moving from marketing-driven to scenario-driven. AI agents have begun to be mass-produced and installed on cars, bringing a new experience, and big models of smart cockpits will become the key technology for the leap in experience.
