Geely Full-Domain AI 2.0 Ushers a New Era of Intelligent Vehicles

On Jan. 5 in Las Vegas, as global attention converged on CES 2026, Geely Auto Group unveiled a major technological leap with the debut of Geely Full-Domain AI 2.0, signaling a decisive shift in how intelligence is embedded across vehicles. Anchored by the global introduction of the World Action Model (WAM), the company demonstrated how artificial intelligence can evolve from isolated features into a cohesive, learning-driven automotive system.
With this announcement, Geely positioned intelligence not as an add-on but as a foundational capability. Vehicles are no longer limited to executing predefined commands; they can now form judgments, anticipate outcomes, and continuously refine their behavior through experience. Alongside this milestone, Geely revealed the English brand name and software identity of its advanced driver-assistance system, G-ASD (Geely Afari Smart Driving), while showcasing new energy vehicles and core technologies that reflect its ambition to lead globally in intelligent mobility.
Geely Full-Domain AI 2.0 Redefines Automotive Intelligence
Geely’s first-generation full-domain AI framework, introduced in 2025, marked the company’s entry into AI-integrated automotive systems. The evolution to Geely Full-Domain AI 2.0 represents a deeper transformation, shifting from basic environment interaction toward seamless human–machine collaboration.
At the center of this transformation is the World Action Model, designed to overcome long-standing challenges caused by fragmented data and isolated control models across driving, cockpit, chassis, and power domains. By unifying these domains, WAM functions as a vehicle-level general brain capable of understanding complex contexts and coordinating actions across systems in real time.
This architecture integrates four tightly connected elements:

• A multimodal large language model that interprets instructions and plans high-level tasks

• An execution module that converts intent into precise vehicle controls

• A predictive world model that simulates physical outcomes within milliseconds

• A value function trained on human driving data to evaluate and select optimal actions

Together, these components allow the vehicle to learn, self-correct, and evolve through continuous feedback.
Agent-Based Architecture Powers Geely Full-Domain AI 2.0
Two structural advances define this new AI generation. The first is agent-based orchestration through a “1+2+N” framework. A central vehicle-level agent coordinates the system, supported by domain agents for smart driving and cockpit functions, while multiple specialized agents manage areas such as energy, chassis, and body control. This structure enables a single voice command to trigger complex, cross-domain collaboration.
The second advance is the consolidation of perception, cognition, memory, and decision-making into shared technical engines. These engines are accessible across domains, significantly improving coordination efficiency and accelerating system iteration without redundant development.
Evolution of the Eva Intelligent Agent
Originally introduced in 2025 as a mass-deployable emotional AI agent, Eva has evolved into a vehicle-level perception and decision hub under Geely Full-Domain AI 2.0. The upgraded agent now combines memory, simulation, and reasoning to deliver more natural and context-aware interactions.
Eva’s development focuses on three dimensions. It is becoming more human-like through wake-word-free voice interaction and adaptive memory that learns user preferences over time. It is more intelligent by interpreting vague, real-world tasks and translating them into executable plans. It is also more useful, integrating cabin and driving functions so users can delegate entire journeys while Eva coordinates mobility and connected ecosystems beyond the vehicle itself.
This progression transforms in-car intelligence from a reactive assistant into a central orchestrator of mobility experiences.
G-ASD Demonstrates the Strength of Geely Full-Domain AI 2.0
Alongside Eva, Geely introduced G-ASD, its next-generation smart driving system built end to end on the World Action Model. G-ASD applies an AI agent architecture that unifies perception, planning, and execution, positioning it among the most advanced driver-assistance systems globally.
Its technical foundation rests on three pillars:

• Extremely rich model architecture, including large-scale multimodal and generative AI models that enhance perception, reasoning, and driving naturalness

• Massive datasets drawn from millions of vehicles and billions of kilometers, reinforced by extensive safety data to improve robustness and reliability

• High-performance hardware platforms featuring multi-lidar configurations and high-compute chipsets designed with redundancy for higher-level automation

Based on this platform, Geely has outlined a clear progression from L2 to L4 automation, with highway-level L3 and low-speed L4 functions targeted for deployment in 2026, subject to regulatory approval.
Strategic Implications for Geely’s Global Ambitions
Through Geely Full-Domain AI 2.0, the WAM framework, the Eva agent, and G-ASD, Geely has signaled a strategic shift beyond conventional automotive competition. The company is positioning itself as a full-spectrum AI enterprise, integrating intelligence from cloud to vehicle to ecosystem.
With cumulative group sales exceeding three million units in 2025 and strong growth in new-energy vehicles, Geely’s roadmap for 2026 reflects confidence in scaling intelligent technologies across its portfolio. As new models roll out under this AI-driven architecture, the company aims to redefine global mobility through systems that are safer, greener, and fundamentally more intelligent.