Why I'm Not Optimistic About Apple's China AI: Qwen Base Model and iOS Ecosystem Adaptation Challenges

After much anticipation, Apple Intelligence for China finally opened its developer Beta test to users in mainland China in the early hours of November 4, and is expected to be fully rolled out with the official release of iOS 27.3 in December. The underlying large model will use a combination of Apple's local model and Alibaba's Qwen, a point that has long been confirmed by announcements from both sides.

I am a loyal Apple user. I cannot imagine switching my phone from the iOS ecosystem to the Android ecosystem. But that does not mean I am optimistic about Apple Intelligence in China. To be precise, I am not optimistic about Apple Intelligence's prospects worldwide, and especially not about its prospects in mainland China—if it turns out I am wrong, I will be very happy.

Apple Intelligence will struggle to fully adapt to the domestic mobile app ecosystem;

iOS first-party apps have insufficient usage rates in China, making it hard to leverage Apple Intelligence's advantages;

The efficiency of the local model plus third-party model approach still needs further observation.

Let me start with the first point. Actually, even in the United States, Apple Intelligence's adaptation to third-party apps is insufficient. The third-party apps that are currently adapted to a relatively deep degree include WhatsApp, Audible, Microsoft Outlook, and so on—among them, only WhatsApp is a genuine top-tier app. Apps under Google and Amazon, as well as other apps under Meta, have varying degrees of integration with Apple Intelligence, and they have never promised any support.

In China, this is even more the case. Given the partnership between Apple and Qwen, we can expect that perhaps Alibaba-affiliated apps will be relatively open. But what about Tencent-affiliated apps, ByteDance-affiliated apps, and other major companies? For one thing, ByteDance very much wants to integrate software and hardware. The Doubao phone did not become popular not because ByteDance did not want it to, but because the conditions were not yet ripe. Given the style of ByteDance's management, I find it hard to imagine ByteDance-affiliated apps opening their arms to Apple Intelligence and integrating deeply with it.

China's mobile internet is inherently an ecosystem in which major companies are closed off from one another and divided into rival camps. This problem troubles not only Apple but also domestic Android manufacturers. Since more than a decade ago, countless hardware manufacturers have fantasized about "hardware as the entry point" and "capturing users through smart devices," but almost all of them were defeated by internet giants. In this process, Apple has relatively high autonomy, but by no means complete autonomy. It can only use a highly centralized oversight system to push third-party apps to comply with its rules, but it cannot make them obey its orders in everything.

Assuming it cannot open up the third-party app ecosystem, Apple Intelligence's main battlefield then becomes iOS first-party apps. The sad part is right here: iPhone penetration in China is high, but Mac penetration is insufficient, and as a result, the usage rate of Apple's entire suite of first-party apps is insufficient. Don't misunderstand—I am not saying that no one in China uses a Mac, only that Mac penetration in China, especially in the productivity field, lags behind that of the United States and other developed countries. (I could have omitted this sentence, but I know that if I did, there would definitely be contrarians who would make an issue of it. This is why I increasingly prefer dealing with AI rather than with real people!)

Pages, Numbers, Keynote... These Apple first-party office apps have a user base in China far behind Microsoft Office and Kingsoft WPS. Apple Notes is used by almost no one in China. (Note: I know some people use it—I use it myself—but few people use it, so to guard against contrarians, I still have to add this parenthetical. This once again proves that I am writing as a human, not as AI!) FaceTime's niche in China has been almost completely taken over by WeChat video chat. I believe that in the early days of Apple Intelligence in China, the most frequently used feature will be photo editing, because everyone still uses the Apple Photos app.

Apple's smart translation is a very powerful feature, deeply integrated with Messages, Phone, FaceTime, and AirPods. However, there is a big problem: at this stage, translation can only be done through AirPods, and users who own AirPods are, after all, a minority. The second big problem is that this kind of smart translation is obviously not suitable for WeChat voice calls, Tencent Meeting, Feishu, and DingTalk voice conferences, yet most voice conversations in China actually take place in these scenarios. We can certainly hope that these major companies will have a change of heart and fully integrate Apple Intelligence's translation feature, but frankly, we might as well hope that aliens will descend tomorrow and solve all the problems on Earth.

We can also copy and paste content from other apps into Apple first-party apps—for example, paste text from WPS into Pages and let Apple Intelligence analyze and summarize it. The problem is, why do that? In such a scenario, what advantage does Apple Intelligence have compared with Doubao, DeepSeek, or the Qwen app? To be honest, it is not even as good as WeChat AI, because most Chinese users certainly spend more time using WeChat every day than using Apple first-party apps. (Note: I can smell the contrarians. They are planning to show up in my comment section, but I will not release their argumentative remarks.)

As for the combination of Apple's on-device model plus third-party large models, this model is actually nothing new for Apple AI—it's just that the domestic provider has been swapped to Qwen. To be honest, the problems with this model have always been significant: complex requests get redirected into third-party apps to be completed, which is a major blow to user experience. Two or three years ago, we could still hope that Apple's model would one day grow strong enough to handle all tasks; now everyone can see clearly that Apple is unlikely to become part of the first tier of large models, because the first tier is already too crowded.

Apple's model has passed review in China, which is a good thing and also a prerequisite for the launch of Apple AI in China. However, the problem of on-device model plus third-party large models has always been there and has not disappeared; the China version also adds a cross-border collaboration problem. There is a lot more I want to say about this issue, but given space constraints, I will not elaborate here.

PS: The image search feature of Apple AI in China will connect to Baidu Image Search, which shows that Baidu has not been eliminated from Apple's AI ecosystem. But what role will it ultimately play? Specifically for image search, will Apple offer more choices to users in China in the future? These questions will probably only be resolved when the full rollout happens in December.

The success of the iOS ecosystem in China is actually the result of "half of its advantages being utilized and the other half not":

The half that has been utilized is iOS's strict management of third-party app permissions, its protection of user data, and the good adaptation mechanism brought by hardware standardization. As long as you have an iPhone, you can enjoy these advantages, and the iPhone's share of Apple's revenue in China has consistently been higher than its share of Apple's revenue in the United States.

The half that has not been utilized is iOS's first-party apps and its integration with the Mac. It must be admitted that in China, in the past, present, and foreseeable future, Windows PCs dominate; I myself have always used Windows and cannot enjoy the powerful integrated Apple ecosystem. Every year a small number of users switch to Mac, but after all these years, the accumulated total is still far from enough.

This does not mean I am pessimistic about the future of Apple phones in China. The iPhone 16 series and 17 Pro series have just set sales records in Greater China, and even if Apple AI in China completely fizzles, it should not have much impact on the market share of the iOS ecosystem. Almost no iOS users will switch to Android just because Apple AI is not good—if they really cared, they would have done so in the past two years.

This article comes from the WeChat public account "Internet Phantom Thief Group" (ID: TMTphantom), author: Phantom Thief Group Leader Pei Pei, published by 36Kr with authorization.