Great Wall Motor Reveals 30% NOA User Retention, Nearly 90% Reuse Rate of In-House ADAS Stack, and Details Yuan Platform and Vehicle Intelligence Strategy

At a recent forum, She Shidong, Deputy General Manager of Intelligent Products at Great Wall Motor, revealed that user engagement with its Navigation on Autopilot (NOA) feature has surged from single-digit percentages to over 30% in the past two years. In terms of in-house developed intelligent driving technologies, the reuse rate of engineering and data pipelines has reached nearly 80%–90%. She emphasized that Great Wall’s newly launched “Guiyuan Platform” is not merely a technical platform but an integrated vehicle design philosophy centered on “one vehicle, multiple powertrains,” enabling adaptation to diverse global market demands—including plug-in hybrids, battery electric vehicles, gasoline, diesel, and conventional hybrids. The platform features a native AI full-stack architecture with two dedicated computing domains: intelligent driving and cockpit, delivering combined computing power at the 1,000 TOPS scale. Specifically, the intelligent driving domain leverages NVIDIA’s Thor chip, offering over 700 TOPS of computing power and integrating a Vision-Language-Action (VLA) large model. Meanwhile, the cockpit domain utilizes Qualcomm’s flagship chip and deploys the industry’s first cabin-space VLA model, capable of in-cabin scene perception, occupant recognition, and generative proactive services. The platform unifies control over all underlying actuators through a “left-brain/right-brain” coordination mechanism, breaking away from traditional “siloed” development approaches and enabling capability-based encapsulation and cross-domain interoperability among subsystems. User interaction is centrally managed by the AI agent “Xiao Wei,” delivering a seamless, integrated mobility experience. In advancing its intelligent driving strategy, Great Wall adopts a dual-track approach combining in-house R&D with supplier collaboration. For flagship models, the company plans to incorporate actuators such as steering and powertrain systems into end-to-end model co-training. Mid-tier, cost-effective solutions (with 100–200 TOPS computing power) will gradually transition entirely to in-house developed systems, covering highway and select urban scenarios. High-end and flagship vehicles will remain open for algorithm co-development with leading external suppliers. She identified three major challenges in vehicle-wide AI integration: insufficient edge-side computing power limiting large model deployment, lack of mature solutions for 3D in-cabin spatial modeling, and difficulty in effectively injecting automotive domain-specific knowledge into models. To address these, Great Wall is intensifying R&D efforts in understanding in-cabin spatial relationships and pre-training foundation models, with plans to equip its next-generation cockpit system with over 300 TOPS of edge-side computing power to support millisecond-level perception and real-time inference. Reflecting on Great Wall’s intelligent transformation journey, She outlined three phases: Phase One focused on platform consolidation, unified software experiences, and establishing high-frequency OTA capabilities; Phase Two involved deep user co-creation and B2C transformation through over 1,000 user interviews and more than 60 instances of 48-hour ride-along research; and the current Phase Three—the “Agent Era”—centers on AI agents evolving through four relationship stages: “meeting, knowing, loving, and companionship,” ultimately aiming to deliver truly personalized “thousands of faces for thousands of users” services. Additionally, Great Wall plans to launch next year a specialized intelligent driving system calibrated for off-road and off-road-oriented vehicles, with enhanced perception capabilities for longitudinal road undulations and 3D obstacles.