3 6 Ke·2025-10-15 10:15Core Insights - The automotive industry's driver assistance systems are rapidly transitioning to AI, with leading new players approaching the parameter scale of many AI large models [1] - Xpeng Motors and Li Auto are both developing in-car large models with parameters reaching at least 7 billion, indicating a significant shift towards AI-driven solutions in the automotive sector [1][5] Summary by Sections Xpeng and Li Auto's AI Strategies - Xpeng's in-car large model is distilled from its cloud-based "Xpeng World Foundation Model," addressing limitations in computing power and storage for in-car deployment [2] - By the second half of 2024, Xpeng plans to transition to a cloud-based model with a foundational parameter count of 72 billion, set to be unveiled at the upcoming AI Technology Day [2] - Li Auto's current in-car model has over 4 billion parameters, with plans to exceed 7 billion next year following the deployment of its self-developed driving chip [5] Technical Developments - Xpeng's "Turing" AI driving chip, launched in June, is designed for AI needs and can handle models with up to 30 billion parameters, showcasing significant advancements in hardware [4] - Li Auto's approach involves a dual-system model combining fast and slow systems, with a focus on the VLA model, which has become the mainstream technology in embodied intelligence [6][5] Industry Trends and Comparisons - Despite the push for larger models, companies like Tesla have achieved superior driving assistance performance with fewer parameters, suggesting that model size does not directly correlate with effectiveness [7][8] - The emphasis on end-to-end technology, which mimics human driving behavior, contrasts with the logic reasoning capabilities of large models, indicating a potential misalignment in resource allocation [8] Motivations Behind AI Model Adoption - The drive for larger AI models is partly due to companies like Li Auto redefining themselves as AI enterprises, viewing vehicles as applications of artificial intelligence [9] - The marketing impact of AI models, influenced by the success of Chat-GPT, has led companies to promote their AI capabilities as a competitive advantage, although the primary goal should remain enhancing driving assistance experiences [11]