> ## Content Index > Fetch the complete content index at: https://www.betteratcoding.com/llms.txt > Use this file to discover other available public pages before exploring further. # Apple built a China-only model with Alibaba ahead of an iOS launch - URL: https://www.betteratcoding.com/trending-news/apple-built-a-china-only-model-with-alibaba-ahead-of-an-ios-launch/ - Published: 2026-09-01T03:00:56.000Z - Updated: 2026-09-01T03:00:56.000Z - Author: Zacarias Ripoll Cid - Tags: trending-news Reuters reported on August 14, 2026, that Apple has trained a large language model just for China, with support from Alibaba. The Verge, 9to5Mac, and MacRumors all carried the same core story. Reuters cited three people familiar with the matter, unnamed because the information is sensitive and not public. Apple did not immediately comment. Alibaba did not fill the gap in the same reports. This is a change in strategy, not a small vendor swap. Until now, Apple’s pattern in China was to bring generative AI to devices by using Chinese models, because the US models it uses elsewhere are not available there. OpenAI’s ChatGPT and Anthropic’s Claude are the examples Reuters named. In other markets Apple pairs its own stack with those US models. In China that pairing is not on the table. The new report says Apple did not only wrap a partner model this time. It trained a China-focused model of its own, in partnership with Alibaba, using Alibaba’s help in training. Reuters said Apple’s training of its own China-specific model had not been reported before. Context from earlier this year still applies, and it is easy to mash the two layers together. In February 2025, Alibaba chairman Joe Tsai said at the World Governments Summit in Dubai that Apple would use Alibaba’s AI to power phones in China, after talking to several firms and choosing Alibaba. In July 2026, Reuters reported that Alibaba’s Qwen model would be incorporated into the China version of Apple Intelligence on compatible iPhone, iPad, Mac, and Vision Pro hardware, and that Baidu technology was also in the mix. Apple also briefly published a China support guide about connecting Qwen to Siri and Writing Tools on Mac, then pulled it in less than a day, according to MacRumors. What Reuters could not say on August 14 is how the new Apple-trained model will sit next to Qwen and Baidu. That part is still unclear. You should read the news as “Apple trained its own China model with Alibaba,” not as “Qwen is cancelled.” Both things can be in the plan until Apple says otherwise. The regulatory track is clearer. Last month the Cyberspace Administration of China, the internet regulator, registered Apple’s on-device generative AI service. In China, models have to be registered and cleared before a public release. That registration was the main gate Apple had not yet passed. Sources told Reuters that Apple Intelligence should arrive in China in the coming months, after an iOS update. The Verge noted that a custom model would, if the reporting holds, make Apple the first US company approved to offer a proprietary AI model in China. MacRumors called this a dual-track approach: keep working with local partners, and also own a model that can be registered as Apple’s. The Verge framed it as a rare US-China AI partnership at a time when political tension around AI is up, not down. Those are interpretations of the same facts: a US device maker trained a local model with a Chinese firm in order to ship on-device features in a market that will not take ChatGPT. Product timing elsewhere is a separate calendar. 9to5Mac noted that Apple’s overhauled Siri AI is scheduled to launch with iOS 27 this fall where it is available, starting with some English variants, with EU iPhone and iPad availability delayed. China’s date is still “coming months after an iOS update,” not a public day on the calendar. Do not collapse those launches into one event. China is its own pipeline because the models and the regulator are different. For a frontend engineer, the useful factual takeaway is about shipping under a closed door. The usual model was blocked. Apple used a local partner, trained a model that can be registered, and waited for the regulator before promising a device rollout. The user-facing moment, if it happens, will look like an iOS update with Apple Intelligence features on Chinese hardware. Underneath that update, according to this reporting, is a custom model plus Qwen and possibly Baidu, with the exact split not disclosed. A few claims should stay hedged, because the sourcing is unnamed and Apple has not confirmed. “Trained with Alibaba’s support” is what the three people said. “First US company with a proprietary model approved in China” is Reuters’s reading of the approval path, not a quote from Apple. “Launch in the coming months” is also from those people, tied to an iOS update, not a keynote date. If you write this up internally, keep those hedges. Unnamed sources plus no comment means the story can still move. What we can say with less hedge is the public track around it. Tsai confirmed a commercial AI partnership in February 2025\. Reuters described Qwen and Baidu in the China Apple Intelligence plan in July. The regulator registered the service last month. August 14 adds a new piece: Apple also trained its own China model with Alibaba. The open question is how those pieces are wired in the shipping product. Devices in scope for the China Apple Intelligence plan, from the July reporting, are compatible iPhone, iPad, Mac, and Vision Pro. On-device is the phrase The Verge used for the registered service. On-device matters in China for the same reason it matters anywhere with data rules: more of the feature can run on the phone instead of a US cloud that is not allowed to serve that market. **Short take:** this is a shipping story more than a model-wars story. Apple could not bring its usual US models into China, so it built and registered a local path with Alibaba. Until Apple talks, treat the custom model, Qwen, and Baidu as pieces of one unreleased stack, not as three competing headlines.