On April 25, Autowise.ai convened a technical press conference at the Beijing Auto Show. During the conference, CEO Guang Zhou revealed that significant advancements in multimodal large model capabilities are anticipated by early 2026. He emphasized that the large model autonomous driving approach represents a substantial leap forward compared to the previous generation of technology. In contrast, small model autonomous driving systems are prone to a "seesaw effect," where performance in one area may compromise safety in another, ultimately failing to provide comprehensive safety coverage across all driving scenarios.

Autowise.ai has now fully committed to the large model route, with the ambitious goal of achieving thousand-kilometer-level MPCI (a metric likely referring to a measure of autonomous driving performance or safety) within this year. Additionally, the company introduced Chuang Ruan, a former core researcher of multimodal technology at DeepSeek, as its new Chief Scientist. Ruan highlighted that Autowise.ai is transitioning from a fragmented approach using multiple small models to a unified large model architecture. This architecture is built upon a foundational model that differentiates into three specialized models for driving, analysis, and evaluation. Notably, the iteration cycle for a single model has been dramatically reduced from over 100 hours to just over 10 hours, showcasing the efficiency gains of this new approach.