Li Auto Unveils MindVLA-o1 Foundation Model

[Technology Launch] At NVIDIA GTC 2026, Li Auto introduced MindVLA-o1, a foundation model for embodied intelligence in autonomous driving.

Core Development: Advancing from 'Monkey Driving' to Human-Level Autonomous Driving

Zhan Kun explained that MindVLA-o1 is built upon a native multimodal approach within a VLA architecture, unifying spatial understanding, reasoning, and behavioral decision-making, while integrating an implicit world model to achieve general intelligence in the physical world.

Industry Impact: Foundation Models Become the New Benchmark for Intelligent Driving

In March 2026, IM Motors and Zhuoyu Technology also launched their respective Level 4 foundation models, signaling the industry’s entry into a new phase where 'no foundation model, no top-tier status.' Edge-based physical AI brains are accelerating toward mass production and deployment.