Zheng Quan Shi Bao·2025-06-17 04:25Core Viewpoint - Alibaba's Tongyi Qianwen team has launched and open-sourced the Qwen3 series of 32 MLX quantitative models, which are designed to be compatible with Apple's MLX framework, paving the way for the implementation of a Chinese version of "Apple Intelligence" [1][6][9] Group 1: Product Development and Features - The Qwen3 series has been optimized to fit the MLX framework, offering various quantization versions (4bit, 6bit, 8bit, and BF16) to ensure compatibility across Apple's product line, including Mac and iOS devices [3][5] - The models range in size from 0.6 billion to 235 billion parameters, addressing the computational power differences across Apple devices, allowing low-quantization versions to run smoothly on devices with limited memory [5][6] - MLX, designed specifically for Apple's chips, provides a PyTorch-like API and supports various programming languages, enhancing efficiency and performance for AI model deployment on Apple devices [5][6] Group 2: Market Context and Strategic Importance - Apple's smartphone market share in China has been under pressure, with a reported 25% drop in shipments in Q4 2024, prompting the need for localized AI solutions to boost sales [8] - The collaboration between Apple and Alibaba is seen as crucial for Apple to meet local compliance requirements and enhance its AI capabilities in the Chinese market [7][9] - The integration of Qwen3 into Apple's ecosystem is expected to improve AI functionalities, offering more tailored services for Chinese users, such as enhanced voice interaction and personalized recommendations [8][9] Group 3: Regulatory Challenges - The partnership between Apple and Alibaba faces scrutiny from both U.S. and Chinese governments, with ongoing reviews impacting the rollout of AI features in China [9] - Despite expectations for the launch of AI functionalities in iOS 18.6, recent developments at Apple's WWDC 2025 did not indicate progress in regulatory approvals for the Chinese market [9]