Leapmotor wanted to turn its upgraded smart cockpit into a more natural AI interaction surface, but traditional fixed Q&A could not support open conversation, vehicle-how-to guidance, image generation, or broad knowledge queries. The automaker needed an architecture that could combine large-model dialogue, enterprise vehicle knowledge, internet knowledge, and visual generation inside one in-car assistant.
Reported impact
Alibaba Cloud helped Leapmotor move from rigid voice commands to a more expressive AI cockpit experience. The production system now supports conversation, vehicle knowledge Q&A, route and feature guidance, and voice-driven image generation on one extensible model foundation, giving the automaker a faster path to add new in-car AI experiences while lowering the cost and complexity of expanding cockpit use cases.
Leapmotor used Alibaba Cloud Bailian as the application layer for a large-model cockpit assistant rather than bolting one feature onto its existing voice stack. Qwen models power open-ended multi-turn voice interaction so drivers can move beyond fixed command trees into natural chat, general knowledge questions, and richer help flows. The assistant retrieves from both Leapmotor's own vehicle knowledge base and a broader internet knowledge base, which lets the same interaction surface answer product-specific questions, explain vehicle functions, and support route-planning or media requests with more context. For visual features, the voice assistant can call Wanxiang through the cloud model layer to generate wallpapers from speech in seconds. Bailian's unified model foundation also gives Leapmotor an extensible application architecture, so new cockpit scenarios can be added and iterated without rebuilding the serving stack each time.
Voice requests enter the Bailian application layer, where Qwen models handle dialogue and retrieve context from Leapmotor's enterprise knowledge base and broader internet knowledge. For visual requests, the assistant invokes Wanxiang through the same cloud stack and returns generated cockpit wallpaper back to the in-car interface.
Leapmotor also uses voice to call Wanxiang for near-instant wallpaper generation and to support questions about vehicle functions and route planning. (original: 零跑采用语音助手调用云端千问系列大模型，帮助用户通用语音调用万相实现文生图换壁纸，实现秒级作图... 支持语音查找如何使用汽车功能、规划路径等功能。)
Leapmotor said its latest OTA upgrade added a voice large-model function in the cockpit for chat, basic knowledge Q&A, and text-to-image scenarios. (original: 零跑汽车已对 OTA 功能完成大规模升级，携手阿里云首次在座舱场景中增加“语音大模型”功能，用于聊天、基础知识问答、文生图等场景。)
Qwen models let users hold open-ended multi-turn voice conversations and combine enterprise and internet knowledge bases for richer answers. (original: 阿里云千问大模型帮助零跑汽车改变了传统的固定形式的问答模式... 同时结合企业知识库和互联网知识库，满足用户多元化的需求。)
The company said Bailian gave it an open and extensible large-model application architecture that lowers the innovation threshold and cost of expanding cockpit AI scenarios. (original: 基于百炼平台，零跑汽车构建了开放、可扩展的大模型应用架构... 降低大模型应用的创新门槛与成本。)
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Alibaba Cloud Sample Jun 10, 2026 Geely Auto Geely's user-data systems were becoming too complex for manual operations after multi-cloud growth, cross-site deployments, and the addition of Zeekr into the group. Topology was opaque, logs and metrics stayed siloed,… target service objective is 1 minute to detect, 5 minutes to locate, and 30 minutes to recover Application Real-Time Monitoring Service (ARMS) · Cloud Monitor 2.0 · Log Service SLS Alibaba Cloud Sample Oct 31, 2025 Xpeng Motors (小鹏汽车) Xpeng needed one secure AI platform for employees instead of a growing collection of disconnected tools. Internal knowledge was sprawling across many departments, employees wanted generative AI for writing, reporting,… covers 44 service-desk knowledge domains AI Safety Guardrails · API Gateway · Bailian Alibaba Cloud Sample Oct 17, 2025 China FAW Group China FAW needed one AI layer that could make unstructured enterprise knowledge searchable, automate employee business handling across many internal systems, and shorten management reporting cycles that still depended… BI report development cycle reduced from 7 days to 5 seconds Alibaba Cloud Bailian · Domain-specific agents · Enterprise knowledge vector database 