Apple Expands Siri AI Overhaul With Public Beta and Wins Regulatory Approval for China Launch

According to a report by GeekPark, Apple Intelligence for China has finally been confirmed to launch, with the underlying model being Alibaba's Qwen. This choice is not accidental: Apple faces strict AI regulatory requirements in China, ensuring model content compliance and data localization. As a leading domestic large model, Alibaba's Qwen has deep expertise in Chinese language understanding, multimodal capabilities, and compliance, and Alibaba Cloud's infrastructure can meet Apple's high standards for privacy and performance.

In contrast, although emerging models like DeepSeek have impressive technical metrics, their commercial maturity and regulatory adaptation experience are not yet on par with Alibaba. Apple's decision essentially provides the strongest endorsement for Qwen, potentially accelerating its penetration into the enterprise market.

The same report notes that shared mobility platforms Meituan, Qingju, and Hello have collectively raised prices. This phenomenon is driven by multiple factors: operating costs for shared bikes and e-bikes continue to rise, including vehicle depreciation, maintenance labor, and battery replacement; meanwhile, the industry's competitive landscape has stabilized, with leading companies no longer relying on subsidies to grab market share, instead pursuing profitability.

The price increases vary by city and vehicle type, but generally range from 20% to 30%. For users, higher short-distance travel costs may push some toward public transit or walking; for platforms, this is a critical test of user stickiness and pricing power. If order volumes do not drop significantly after the hikes, it could signal a new round of profitability for the industry.

The report also reveals that DeepSeek is preparing for an IPO. As an AI company known for open-source models and efficient training, DeepSeek gained widespread attention with models like DeepSeek-V2. However, Apple's choice of Qwen over DeepSeek as its AI partner in China may weaken its commercial appeal.

The success of the IPO will depend on whether DeepSeek can find a differentiated position in the B2B market, such as building advantages in vertical industries or small and medium-sized enterprises. Additionally, its open-source strategy has won over the developer community, but converting that into sustainable revenue remains a challenge.

Nvidia CEO Jensen Huang confirmed that the next-generation computing platform Vera Rubin is already in production and progressing smoothly. This marks a new phase in the AI computing power race, with Vera Rubin expected to significantly improve training and inference efficiency, further solidifying Nvidia's leadership in AI chips.

Meanwhile, Google DeepMind CEO Demis Hassabis has called for the US to lead the establishment of a global AI regulatory body. This proposal reflects industry concerns about AI safety and governance, but actual progress faces challenges from geopolitics and coordination of national interests. Google itself continues to push AI applications; on the 25th anniversary of its image services, AI Overviews added the Nano Banana image generation feature, further lowering the barrier to image creation.