Three sources familiar with the matter told Reuters on August 14 that Apple has already completed training a large language model for the Chinese market, developed in collaboration with Alibaba. The move marks a major pivot from Apple's original plan to rely entirely on domestic partners like Alibaba and Baidu for AI functionality in China.
Apple is now pursuing a "dual-track" strategy: using its own lightweight, on-device model for core AI tasks, while Alibaba's Qwen handles heavier cloud-based workloads like long-text processing and multimodal generation. In the US, Apple's AI strategy depends on Anthropic's Claude and OpenAI's ChatGPT. In China, those models are not available—so Apple is making its own.
A Support Document Leaked. Then Apple Deleted It.
The change happened in stages. In February 2025, Alibaba chairman Joe Tsai confirmed Apple had chosen Alibaba for AI cooperation in China. The original plan: Apple would integrate Qwen directly into Apple Intelligence for Chinese users. The service completed China's generative AI service registration on July 15—a key regulatory step. Apple would be the first foreign company approved to offer a proprietary AI model in China.
But on August 8, something unusual happened. A support document titled "Using Qwen with Apple Intelligence on Mac" appeared on Apple's Chinese website, detailing exactly how Siri and writing tools would work with Qwen. Within a day, Apple removed it—without explanation.
The deletion wasn't a mistake. It was a signal that the plan had changed.
Behind the scenes, Apple had been working on its own China-specific model with Alibaba's support, rather than simply integrating Qwen directly. The document that leaked was from the old plan. The new plan is different: Apple is building and training its own model for Chinese users.
Apple Runs the Core. Alibaba Runs the Cloud.
The current strategy splits responsibility. Apple's own lightweight model handles core system-level AI interactions—device-side inference, real-time responses, and the core Apple Intelligence experience. Alibaba's Qwen manages the heavier tasks—long-text generation, complex multimodal processing, and workloads that require cloud-scale computing.
It's a dual-track approach that balances compliance with control. Apple stays responsible for user experience, data handling, and core interactions. Alibaba provides the compute infrastructure, Chinese-language training data, and regulatory compliance support. By keeping the core model in-house, Apple retains control over what Chinese users experience—and how it adapts to local preferences and regulations.
Apple Intelligence is expected to launch in China in the "coming months" with an iOS update. The registration process has already cleared the largest hurdle.
The "One Model Fits All" Era Is Over
Apple's China move marks a real shift in how big tech handles cross-border AI. The "one model fits all markets" approach is over. Apple's strategy—build a core model in-house, use local partners for infrastructure, and tailor the experience to each country's regulatory environment—is likely to become the default for US tech companies operating globally.
For US AI providers, the message is clear: if you want to operate in China, you need local partners. Apple chose Alibaba for its infrastructure, compute capacity, and regulatory experience. The deeper implication is that Chinese cloud providers and AI companies are becoming essential infrastructure for global tech companies that want to operate in the world's second-largest economy.
P.S. The "secret" model isn't secret to regulators—Apple completed its generative AI service registration in July, and Apple Intelligence's launch appears imminent. What's notable is the shift: from "Apple will use Qwen" to "Apple trained its own model with Alibaba's help." That's a much larger commitment to local engineering than was previously understood.
Frequently Asked Questions
Q: What did Apple do with Alibaba?
A: Apple trained a large language model specifically for the Chinese market in collaboration with Alibaba. This is a departure from Apple's original plan to simply integrate Alibaba's Qwen model directly into Apple Intelligence.
Q: Why did Apple change its strategy?
A: Apple originally planned to use Qwen directly. But Reuters reported that Apple has now trained its own China-specific model with Alibaba's support, rather than simply integrating Qwen. The shift gives Apple more control over the user experience and data handling while relying on Alibaba for infrastructure and regulatory compliance.
Q: What is the "dual-track" AI strategy?
A: Apple's own lightweight on-device model handles core system-level AI interactions. Alibaba's Qwen handles heavier cloud-based tasks like long-text generation and multimodal processing. Apple stays responsible for user experience and core interactions; Alibaba provides compute infrastructure and compliance support.
Q: Did Apple Intelligence already get approved in China?
A: Yes. Apple completed China's generative AI service registration on July 15, 2026. This was a key regulatory hurdle that cleared the path for Apple Intelligence to launch in China.
Q: When will Apple Intelligence launch in China?
A: Apple Intelligence is expected to launch in the "coming months" with an iOS update, according to Reuters.
Q: What happened with the August 8 support document?
A: A support document titled "Using Qwen with Apple Intelligence on Mac" appeared on Apple's Chinese website on August 8, then was removed within a day. It described exactly how Siri and writing tools would interact with Qwen. The deletion signaled that Apple's original integration plan had been replaced.
Q: What does this mean for other US tech companies?
A: Apple's model suggests that operating in China increasingly requires local partnerships and localized products. For US tech companies, the "one model fits all" approach is no longer viable.
Q: Is Apple the only foreign company approved to offer AI in China?
A: Apple is the first foreign company to have its generative AI service registered and approved in China—it is also the first to offer a proprietary AI model with both on-device and cloud components.
Q: How will this affect user experience?
A: Users will see Apple Intelligence integrated into iOS, macOS, and iPadOS—but with features adapted to the Chinese market. Siri and writing tools will be available in China after the update, powered by a hybrid of Apple's local model and Qwen's cloud backend.
