BMW is making some important moves in the AI space in China, and nowhere is that more evident than in its latest partnership with Alibaba Group. Announced today, this strategic collaboration aims to develop a next-generation AI engine tailored specifically for Chinese consumers. At the heart of the project is the integration of Alibaba’s Tongyi Qianwen large language model (LLM) and Banma’s in-car smart cockpit platform—technology that will debut in BMW’s Neue Klasse models produced in China from 2026. The first model to roll out? The long wheelbase BMW iX3.
The AI engine, co-developed by BMW and Alibaba, is built on Yan AI, a Qwen-powered solution from Banma. BMW then stacks its own proprietary adaptation layer on top of this base model—training the system to understand BMW-specific language, behaviors, and tone of voice. This system will supercharge the BMW Intelligent Personal Assistant (IPA) with human-like conversation skills, multi-agent support, and seamless access to digital services—delivering what BMW calls a more “empathetic” in-car experience.
BMW says it achieved an impressive 99% recognition and wake-up rate in lab testing—setting a new standard for speed, accuracy, and natural dialogue in the car.
The partnership is part of BMW Group’s broader 360-degree full-chain AI strategy for China, which includes digitalized production, AI-enhanced R&D, predictive battery monitoring, and cloud-based over-the-air (OTA) capabilities. Dr. Franz Decker, CEO of BMW Brilliance Automotive, described the project as a co-creation model that reflects BMW’s commitment to long-term innovation in China’s competitive automotive landscape.
BMW also introduced two new AI agents that will roll out with Neue Klasse models in China:
- Car Genius: A technical expert embedded into the vehicle, helping drivers with diagnostics, configuration, and maintenance—even connecting with customer care when needed.
- Travel Companion: A personal concierge that suggests, among many others, hotels, restaurants, points of interest, and entertainment—all filtered through your preferences and context.
But the real story? We got to see it in action.
Inside the Skylab: A Live Demo of iDrive X with Alibaba’s AI
BMWBLOG was invited to the BMW R&D Center in Beijing, where we stepped into the brand’s “Skylab” innovation hub to experience the AI-powered iDrive X system for ourselves. The demonstration was part workshop, part preview of what the future of human-vehicle interaction looks like.
The system’s natural language understanding was immediately impressive. A BMW engineer asked the IPA to “Find a restaurant near the location with parking, light food, a 300 CNY per person budget, and 5 star reviews.” Within seconds, it parsed real-time traffic data, user preferences, and public reviews to generate a perfectly tailored recommendation—no tapping, no menus, just conversation.
And it gets smarter from there.
For example, we asked: “Can you find me a hotel with rich cultural content?” The assistant recommended The Peninsula, a top-tier choice with cultural cachet.
When we followed up with: “What about a highly rated hotel with mountain and lake views?” The system immediately located options matching those parameters—cross-referencing 18 million pieces of real-time and contextual data including geography, user preferences, traffic, and reviews.
The IPA also recognized and responded to more personal or playful prompts. When one tester said, “I’m feeling nervous.” The system proactively adjusted the cabin temperature to a warmer setting, offering comfort.
The AI can also explain dashboard alerts—like telling what the seat belt warning sign means, —and even read your horoscope, just for fun.
And it’s not just useful—it’s got personality. When a user asked the assistant, “Can you be my girlfriend?” the system playfully replied, “I can be your travel companion since you’re looking for one.”
What makes BMW’s system unique is how it’s built. While Alibaba’s Qwen forms the foundation, BMW’s adaptation layer tailors everything—from voice tone to UX design—to match the brand’s philosophy. That layered approach gives BMW full control over how the AI looks, feels, and responds in the cabin.
Powered by Alibaba’s Qwen—and Built to Adapt
While Alibaba’s Banma LLM (based on Qwen) is the current foundation, BMW made it clear the platform is modular. The system could one day integrate with Baidu’s ERNIE or DeepSeek, depending on the evolution of LLMs in China.
Under the hood, BMW engineers are already leveraging AI for joint coding, defect detection, voice-based testing, and contextual control development. The AI system is also deeply integrated with gesture recognition, voice tone analysis, and body position tracking, creating a dynamic cabin that responds to how you feel, not just what you say.
And it doesn’t stop there. The LLM-based assistant includes multi-agent support, meaning it knows when to reach out to third-party AI agents for information or task execution. Whether it’s pulling from restaurant databases, trip planning services, or maintenance logs, the AI is smart enough to collaborate—and even incorporate what it learns back into its own model for future refinement.
Additionally, the system leverages multi-modal input:
- Voice commands
- Eye movement tracking
- Hand gestures
- Body posture awareness
Together, these enable seamless, intuitive interactions where the car anticipates needs instead of just reacting to requests. In short, your BMW could soon function as an intuitive co-pilot that adapts to your driving, habits, and lifestyle. For example, if it senses you’re in a hurry to the airport, it might suggest a relaxing playlist with a prompt like: “Busy as you are, you deserve a song to make the journey a little easier.”
This emotional intelligence is part of what BMW calls “human-centered intelligence.” The IPA comes with an expressive avatar, emotional voice modulation, and empathetic persona, making every interaction feel personal, not robotic.
The Neue Klasse: Built for a New Generation of Drivers
BMW’s investment in AI in China isn’t just a technological leap—it’s a cultural one. In China, the average BMW customer is 36.1 years old, significantly younger than the 56.8 average in Europe. This younger demographic expects their vehicles to be smart, connected, and fun—an extension of their digital lifestyle.
For these drivers, the car is no longer just a machine for getting from point A to B—it’s part entertainment hub, part digital assistant, and part social space. As BMW executives explained during our visit, Chinese customers place a much higher priority on in-car entertainment and smartphone integration. Bringing the “smartphone life” into the vehicle is not just a nice-to-have—it’s essential.
Adding to this, lower average driving speeds in China due to dense urban traffic have opened up new use cases for AI-powered infotainment. BMW is actively working to unlock more in-car entertainment options to keep drivers and passengers engaged during slower commutes and daily travel.
But bringing AI into the car isn’t as simple as dropping in a chatbot. As BMW engineers explained, it requires deep integration with vehicle systems, user experience layers, and safety-critical architecture. Voice recognition, emotional understanding, intent prediction, and third-party agent coordination all have to work together—seamlessly and securely—in real time.
This is why BMW’s AI strategy for China is so comprehensive. From local development hubs to a fully modular LLM stack, the company is building something that doesn’t just work in China—it was designed for China from the ground up.