What challenges did BYD face before AI implementation?

Enable a proactive, privacy-preserving in-car assistant that can execute real-world multi-step tasks. Maintain responsive performance under weak or intermittent network conditions. Support low-latency voice recognition and complex AI model operations inside the vehicle.

What was the impact of the AI solution?

In-cabin AI becomes a proactive agent supporting commands and task execution directly in the vehicle. The system enables booking hotels, purchasing attraction tickets, ordering food delivery, and tracking parcels. It extends the in-car experience from cabin control to real-world service execution. The Hongqi implementation improved recognition of ambiguous voice commands and multi-step planning.

Source

Alibaba

2026 Beijing Auto Show: Multiple Chinese automakers integrate Qwen agentic AI into in-cabin systems (edge+cloud)

Use case typeVoice automationUpdated Apr 24, 2026

On the opening day of the 2026 Beijing Auto Show, leading Chinese automakers including BYD, Geely, Li Auto, Changan Automobile, Dongfeng Motor, BAIC, Great Wall Motor, SAIC Volkswagen, and SAIC IM Motors announced integration of Qwen into their intelligent vehicle systems. Select models from these manufacturers will offer Qwen-powered AI services directly within the cabin, enabling users to book hotels, purchase attraction tickets, order food delivery, track parcels, and more. Alibaba Cloud describes an edge + cloud collaborative architecture for smart cockpits: Qwen-Omni runs on the edge to perceive and interpret the physical environment, while cloud-side Qwen agentic AI capabilities understand natural-language commands, decompose intents, plan multi-step workflows, and orchestrate scenario-specific agents for seamless execution. The article also notes an earlier FAW Hongqi 'Lingxi Cockpit' implementation that integrated Qwen agentic AI into an intelligent system on the Hongqi HS6 PHEV for ambiguous voice recognition and complex multi-step task planning. Alibaba says it is also adapting Qwen-Omni to run on NVIDIA DRIVE AGX Thor for in-cabin human-machine interaction.

Organization: BYD
Industry: Automotive
Location: China
Published: April 2026

Reported outcomes

Strategic outcomes

Customer experience & trustTurned the in-car assistant into a proactive service agentOther strategic outcomeEnabled transactional services from the vehicle cabinSpeed & agilityImproved responsiveness under weak connectivity

Primary read

Use case focus

Showing 1 of 1

1Voice automation

Enable a proactive, privacy-preserving in-car assistant that can execute real-world multi-step tasks.
Maintain responsive performance under weak or intermittent network conditions.
Support low-latency voice recognition and complex AI model operations inside the vehicle.

Alibaba Cloud uses an edge + cloud architecture for smart cockpits.
Qwen-Omni is deployed at the edge to perceive and interpret the cabin environment while preserving privacy and responsiveness.
Cloud-side Qwen agentic AI capabilities decompose user intent, plan multi-step workflows, and orchestrate scenario-specific agents for execution.
The implementation leverages Alibaba Cloud infrastructure co-optimized with Qwen and is also being adapted to run on NVIDIA DRIVE AGX Thor.

Technologies

Qwen-Omni Qwen Alibaba Cloud Model Studio Alibaba Cloud infrastructure NVIDIA DRIVE AGX Thor

In-cabin AI becomes a proactive agent supporting commands and task execution directly in the vehicle.
The system enables booking hotels, purchasing attraction tickets, ordering food delivery, and tracking parcels.
It extends the in-car experience from cabin control to real-world service execution.
The Hongqi implementation improved recognition of ambiguous voice commands and multi-step planning.

Architecture

Edge + cloud collaborative architecture: Qwen-Omni multimodal model deployed on the edge for environmental perception and response under weak/intermittent connectivity; cloud-side Qwen agentic AI capabilities handle natural-language understanding, intent decomposition, multi-step workflow planning, and orchestration of scenario-specific agents. The article also mentions adaptation of Qwen-Omni to NVIDIA DRIVE AGX Thor for in-cabin human-machine interaction.

Implementation partners1

NVIDIA

Sources & evidence1

Groundedness: 4/5Type: Blog PostPublished: Apr 24, 2026Publisher: Alibaba CloudEvidence: VendorConfidence: Medium

Primary source

AI-generated summary. Verify important details with the linked sources before relying on this case.

Explore related AI use cases

Customer evidence