Apple Intelligence Set for China: Qwen Powers the Models, Baidu Takes Siri

Speaking to media in Shanghai on October 15, Apple's board executive chairman Tim Cook confirmed what Chinese users had long waited to hear: Apple Intelligence and the related Alibaba Qwen models "have been approved", with a beta arriving in China later this month. "We're glad the first set of core Apple Intelligence features can gradually come online," Cook said.

First features and a two-partner split

According to Cook, the late-October beta will include real-time translation, writing tools and photo cleanup — while Siri AI is explicitly excluded this time. Media reports indicate a "dual-partner" structure for China: Alibaba supplies model capability, while Baidu handles Siri and visual intelligence. Lianhe Zaobao, citing sources, reports the full rollout is planned for December.

Why it took this long

Generative AI services in China must complete algorithm and model filing procedures before launch, so a foreign model cannot simply ship directly — a "local partner + feature split" is effectively the only viable path. For Apple, handing Siri and visual intelligence to Baidu while giving general model capability to Qwen satisfies regulation and preserves pricing and iteration flexibility per module. For domestic models, anchoring to a terminal base measured in billions of users is worth far more than short-term revenue — after open-source ecosystems and API price wars, this is the third front of China's LLM competition: distribution at the entry point.

A giant's arrival is not the endgame, but the start of a reshuffle in how Chinese LLMs reach consumers.

The other side of the same day

That same day Apple closed above a $5 trillion valuation for the first time — yet global tech stocks broadly pulled back overnight on AI funding worries, and amid mostly stronger Chinese ADRs, Alibaba slipped more than 1%. "Good news landed" and "valuation worries" hedged each other in a single session, a sign of how sensitively the market now prices AI partnership narratives.

The lesson for enterprises is equally direct: a hybrid architecture of smart terminals plus cloud LLMs is becoming the default, and cross-model, cross-device orchestration will be a basic skill for enterprise AI over the next two years. NineZenith's Zenith-Act platform sticks to multi-model access and device-cloud synergy, helping enterprises keep their options as the model ecosystem shifts.

(Compiled from public reports by QbitAI, ifanr, Lianhe Zaobao, Guandian, Moomoo, etc.)