Apple Picks Alibaba’s Qwen for Apple Intelligence in China
In short
- Alibaba says its Qwen model will power Apple Intelligence features for iPhone users in China.
- Testing starts in the iOS developer beta on November 4. The public rollout comes with iOS 27.3 in December.
- The money angle: the same Qwen family is open and cheap. Qwen3.8-27B is reported at $2.55 per million output tokens on OpenRouter, half of GPT-6.1 Terra’s $5.
Apple just chose a Chinese AI model to run on iPhones. Not a test. Not a side project. For the Chinese market, Qwen becomes the engine under Apple Intelligence. Why should you care if you live in Bulgaria? Because it shows which model the biggest hardware brand on earth considers good enough.
What happened On October 31, Alibaba said it will integrate Qwen into Apple Intelligence for users in China. Xinhua reported it. Qwen will handle text and image understanding and content generation in features like Writing Tools and Visual Intelligence.
The dates are clear. From November 4, you can test it in the latest iOS developer beta. The public release comes with iOS 27.3 in December.
Who gets it? Users with an Apple Intelligence-capable iPhone bought in China, or with both device location and Apple Account region set to China. It is opt-in. Alibaba says it will not use user requests or answers to train the model.
Two honest notes. Bloomberg reported that Apple itself has not confirmed the date. And reports differ on whether the new Siri is part of the first release, so do not count on it.
This did not come from nowhere. China’s cyberspace regulator listed Apple’s service in its filing records on July 15, and Alibaba confirmed in July that Qwen would power it. October 31 is the first time we get a concrete schedule.
How to use it You cannot switch it on in Bulgaria. But you can use the news.
- Test Qwen yourself. The family is available through API platforms like OpenRouter. Run your own customer replies, product descriptions or summaries through it for a week.
- Building for iPhone users in China? Expect the writing and image features to behave like Qwen, not like ChatGPT. Test your content there from November 4.
- Watch the price gap. When a model sits inside an iPhone, its competitors cut prices to stay relevant.
What it means A Beijing professor, Wang Xiaojie, called it the first time a top global hardware maker uses a Chinese model as the main system-level AI in the Chinese market, as CGTN reports. Apple needs a locally approved model. Alibaba gets distribution on a huge number of phones. Western labs stay out of that market.
For you the lesson is simple: capable models keep getting cheaper, and the cheap ones are now good enough for Apple.
Do the math: a small support bot writes 5 million output tokens a month. On Qwen3.8-27B at $2.55 per million, that is $12.75 a month (about €12), or $153 a year. On GPT-6.1 Terra at $5 per million output, it is $25 a month, $300 a year. You save about $147 a year per bot. Input tokens are extra on both sides, and reasoning models write more tokens per task, so measure your real usage before you switch.
Ivan’s Take
I cannot use this in Bulgaria, and it does not matter. If Apple trusts Qwen with writing on a phone, I can trust it with first drafts and support replies. I would test it on one real task this week and compare the cost with what I pay now. Never pay premium prices for work a cheaper model already does.
Source: CGTN
If it doesn’t save you time or money, you won’t see it here
- Facts, not hypeEvery price and number is checked against the official source
- Always the real priceYou see what it costs before you click
- I say when it’s weakEven when it’s OpenAI or Google
