The Apple–Qwen story flipped on its head within 24 hours.

On August 8, Apple China's Mac User Guide showed a page titled "Use Qwen with Apple Intelligence on Mac," complete with setup steps, Siri invocation, writing tools, and even the fine print: macOS 26.6 or later, with separate guidance for mainland China accounts, locations, and China-market Macs. On the morning of August 9, the manual quietly vanished; the link redirected to the guide's homepage, and the Chinese manual reverted to "Use ChatGPT with Apple Intelligence."

A question immediately surfaced: is the Apple–Alibaba deal in trouble? Hold that thought.

Three Pieces of the Puzzle — Only One Got Pulled

To judge whether this is a "breakup," look at the other two pieces.

Piece one is regulation. On July 15, the Cyberspace Administration published filings for seven new on-device generative AI services — "Apple Intelligence" was explicitly among them. Apple Intelligence's mainland rollout has already crossed a major compliance milestone.

Piece two is Alibaba's own public statement. In mid-July, Alibaba said Qwen would be integrated into Apple Intelligence as an AI capability for mainland iOS, iPadOS, macOS, and visionOS users. In other words, "Apple + Qwen" wasn't a story invented by that webpage.

So deleting one user manual on August 9 doesn't warrant a "deal cancelled" conclusion. What's more telling is the manual's granularity: system versions, account conditions, setup paths, Siri invocation — precision far beyond a marketing line. This looks like product preparation that ran far ahead, while official release timing, system versions, or rollout rhythm are still being adjusted.

Regulatory filing done, partner confirmed, user guide briefly live — Apple China AI is in a highly sensitive final phase. Apple hasn't explained the deletion. That silence, in itself, is the biggest signal.

Qwen Isn't After Users. It's After the System Entry Point

Beyond the "is it dead or alive" debate, what this episode really shows: the battlefield for Chinese LLMs has moved from "whose app has more downloads" to "who gets into the operating system and hardware entry point."

If an AI model sits inside Siri, writing tools, and more system scenarios to come, users never need to open a standalone AI app. For an LLM vendor, that entry point's value dwarfs another icon on the home screen — standing behind Siri means real system-level invocations, not traffic that can be uninstalled at will.

That's why "who Apple picks as its China AI partner" gets so much attention. Apple doesn't just sell iPhones and Macs; it owns a super entry point built from hardware, OS, and app ecosystem. Whoever becomes the AI foundation inside that system gets the conversation rights behind hundreds of millions of Chinese smart devices.

We used to compare Chinese LLMs by benchmark margins. Apple's China AI push offers another lens: Chinese models are already participating in the underlying capability division of the world's top consumer-electronics ecosystem. The future pecking order of Chinese LLMs may hinge on three things: entering super terminals, absorbing billion-user load, and turning model capability into true system-level services.

So that deleted page shouldn't be read as a suddenly dead deal. Quite the opposite — it surfaces a bigger question: the filing is done, the partner has confirmed, the manual even briefly appeared. What the market is really waiting for is Apple finally opening that last door. When that day comes, the prize isn't the chatbot market. It's the AI entry point behind hundreds of millions of Chinese devices.

See you tomorrow.