Great Wall Motor (GWM) is betting on a system-level artificial intelligence strategy to sharpen its competitive edge in overseas markets, according to a keynote speech by the company's chief scientist at the 22nd China Automotive Industry Development (TEDA) International Forum in Tianjin on September 19.
Chen Xiaofeng, vice president of GWM's technology center, said the industry is entering the "second half" of intelligent vehicles, where cars evolve from transport tools into mobile intelligent spaces. He cited data showing that L2-level assisted driving penetration in new cars sold in China jumped from 55.7% in 2024 to 66.1% in 2025, while smart cockpit penetration reached 76.6%, spreading into the mainstream 100,000-yuan (about $14,100) segment.
From feature stacking to a vehicle-wide AI agent
Chen argued that traditional smart features are isolated and reactive, but AI-driven development is moving toward a "vehicle-wide intelligent agent"—a unified interface that can call on underlying capabilities across powertrain, chassis, cockpit, and driving assistance. GWM's approach embeds AI into the operating system itself, rather than adding AI features on top.
The company has built a "1+N" architecture with a central brain coordinating domain-specific agents, and has broken vehicle functions into more than 4,000 MCP atomic capabilities that large models can orchestrate. Chen stressed safety: "Large models can be creative, but the operating system must never hallucinate," citing sandbox isolation and a vehicle-control safety net as safeguards.
The latest generation, Coffee AI OS 4, is already installed in models such as the Wey Gaoshan, the second-generation Lanshan, and the V9X, with more brands to follow. GWM says it has converted over 3,100 historical engineering issues into 240 development specifications and received the industry's first "excellent smart cockpit fluency" certification from TAIER.
From hardware to ecosystem exports
Chen said overseas perceptions of Chinese cars are shifting, with foreign media dubbing them "smartphones on wheels" and German outlets crediting Chinese brands with reshaping local expectations. He argued that Chinese automakers cannot rely solely on cost competitiveness or shallow local adaptation; they must build systematic, reusable AI capabilities.
GWM's export journey, he said, has moved through three stages: hardware-led before 2023, technology-led from 2023 to now, and ecosystem-led starting this year. The company has developed a "1+N+AI" architecture—system base, ecosystem expansion, and AI-driven usage—and already offers models like the Ora 5, Haval H6, and Haval H7 overseas, with more planned.
Chen closed with three proposals: continuous evolution from OS to AI OS, unified standards so Chinese solutions become a global language, and open ecosystems rather than "small yards with high fences." "Chinese cars have moved from following to running alongside to leading—there are no cliffs, only steps," he said, according to D1EV.
