China Tech Today: Apple Intelligence Arrives in China — November 4, 2026

Two years after Apple Intelligence first appeared elsewhere, it has finally reached mainland China — on a Chinese model. Today’s brief is dominated by that arrival, alongside a new Tencent flagship model, a humanoid robot’s full factory shift, and a cooling chip market that has put one of China’s biggest IPOs on ice.

Apple Intelligence enters developer beta in China, powered by Alibaba’s Qwen

Apple began testing Apple Intelligence in mainland China on November 4, with the iOS 27.3 developer beta, TechNode reports. Developers with compatible iPhones can enable the features after installing the update, ahead of a planned public release in December — the payoff of the arrangement Apple and Alibaba announced on October 31.

The first wave covers Writing Tools, Live Translation, Visual Intelligence, the Clean Up photo-editing tool and Genmoji, plus summaries for emails and notifications. Alibaba’s Qwen model is integrated into Siri, Writing Tools and Visual Intelligence, handling more detailed answers, text generation and image understanding; users can tap through to the Qwen app for more advanced requests. Alibaba has said requests and responses will not be used to train its models. The rollout does not include the next-generation Siri AI or the other new Apple Intelligence features announced at WWDC 2026, and it is limited to Apple Intelligence-capable iPhones bought in China, or devices whose location and Apple Account region are both set to China.

Why it matters is the precedent: Apple Intelligence was unavailable in mainland China because it needed regulatory approval, and the solution is a local model inside Apple’s system-level AI rather than Apple’s own — a template other Western tech firms may have to follow, and a distribution channel for Qwen no Chinese rival can match. For Canadian readers, the same iPhone will soon run different AI plumbing depending on which side of the Pacific it was bought on. (TechNode: https://technode.com/2026/11/04/apple-intelligence-enters-developer-beta-in-china-with-qwen-integration/ ; KuCoin/ChainCatcher: https://www.kucoin.com/news/flash/apple-intelligence-beta-launches-with-alibaba-s-qwen-integration)

Tencent launches Hy4, a model that can see and use a computer

Tencent launched Hy4 on November 4, adding visual understanding and computer-use capabilities to its flagship AI model, TechNode reports. The 770-billion-parameter model can interpret charts and documents, reason across multiple images, and operate command-line tools and graphical interfaces, with targeted training for applications such as Blender and CAD software.

Hy4 is available across Tencent’s products and through third-party platforms including OpenRouter, with its weights and a system card due to follow shortly. Standard API pricing is unchanged from the Hy4 preview at $0.834 per million input tokens and $2.501 per million output tokens, with a 50% discount on Tencent Cloud TokenHub through December 31. Tencent has integrated Hy4 into its WorkBuddy and CodeBuddy products, offering a limited free usage quota from November 4 to November 15.

Why it matters: computer use is the next step past chatbots that only answer — models that can actually operate software, fill in a spreadsheet, or drive a design tool. Tencent says development focused on office tasks, financial analysis and game development, evaluated against real workflows. (TechNode: https://technode.com/2026/11/04/tencent-launches-hy4-with-vision-and-computer-use-capabilities/)

Xiaomi’s humanoid robot works a full 8-hour factory shift

Xiaomi has released footage of its full-size humanoid robot working a full 8-hour shift at its EV factory, sorting centre console side covers, Electrek reports. The video, posted by Xiaomi Tech on November 3 as the company collected awards at the World Internet Conference in Wuzhen, shows the robots pulling side covers from bins, carrying them to storage racks and placing them into slots. According to Xiaomi, the robot covered the full day of production needs at that workstation, detected and corrected anomalies on its own, and coordinated with the production line’s control system.

This is the third public milestone in a program Xiaomi has run all year: in March, a humanoid “intern” in its Beijing EV factory installed self-tapping nuts with a reported 90.2% success rate over three hours; by July, Xiaomi said that rate had risen to 98%. Xiaomi also announced trial runs of its open-source Xiaomi-Robotics-1 AI model, pre-trained on 100,000 hours of real-world robot data, in Xiaomi staff apartments.

The caveat matters as much as the milestone: Xiaomi did not say how many parts the robots sorted, what the success rate was over eight hours, or how many times a human stepped in — nor how the robots stayed powered for a full shift. Until those numbers appear, this is a promising demonstration, not proof of unsupervised production. (Electrek: https://electrek.co/2026/11/04/xiaomi-humanoid-robot-8-hour-shift-ev-factory/ ; November 3 brief: https://futuretechpulse.ca/china-tech-today-2026-11-03/)

DeepSeek reportedly plans to train its V5 model on Huawei Ascend chips

DeepSeek plans to use Huawei’s Ascend chips to train its next-generation V5 model, STAR Market Daily reported on November 3, citing multiple industry sources, TechNode reports. Ascend chips have so far primarily supported inference — running DeepSeek models — so using them for training would extend their role into the far more computing-intensive process of building a new model.

Why it matters: training is the hardest test of a chip stack, and the place where US export controls have bitten hardest. If DeepSeek can train a frontier-class model on domestic hardware, it weakens the argument that Chinese AI labs depend on Nvidia’s most advanced processors — and it would validate Huawei’s Ascend line as a training platform, not just an inference substitute. Treat this as a report, not a confirmed plan: neither company has confirmed it, and the report rests on unnamed industry sources. (TechNode: https://technode.com/2026/11/04/deepseek-reportedly-plans-to-train-v5-model-on-huawei-ascend-chips/)

YMTC puts its $4.9 billion IPO on hold as China’s chip stocks slide

Chinese flash-memory chipmaker Yangtze Memory Technologies (YMTC) has put its Shanghai IPO on hold, Reuters reported on November 4, citing a company memo and people with knowledge of the matter. YMTC’s parent, CCSH Corporation, filed for an IPO on Shanghai’s STAR market in August, aiming to raise 33 billion yuan ($4.9 billion), and had planned to kick off strategic share placement in late October and early November. That process is now paused due to “various factors,” according to an internal note seen by Reuters, and the delay could last several months, with a listing now likely in the first half of 2027.

The backdrop is a sharp correction: China’s blue-chip CSI300 index slumped 8.5% in October, its worst monthly performance in more than a decade, and the STAR50 index of chip and AI firms has tumbled 40% from its July peak. Reuters also reports informal regulatory guidance to investment banks to slow down IPOs by AI and chip companies. The delay could complicate expansion plans for China’s largest maker of NAND chips — the memory used in phones, computers and data centres. (Reuters, via WABX: https://wabx.net/2026/11/04/exclusive-chinese-chipmaker-ymtc-puts-ipo-on-hold-as-weaker-markets-weigh-sources-say/)

Tesla’s China-made sales rise 10.3% in October — but fall sharply from September

Deliveries of Model 3 and Model Y vehicles produced at Tesla’s Shanghai plant, including exports to Europe, Asia Pacific and Canada, rose 10.3% year-on-year to 67,829 units in October, Reuters reported on November 3, citing China Passenger Car Association data. That was up from a 5% annual rise in September and a 12th straight month of annual growth — but sales fell 28.9% from September.

The annual figure flatters and the monthly figure tempers: October is typically softer, and the year-on-year streak says more about Tesla’s Shanghai export machine than about demand in any single market. Still, the plant remains a bellwether — when Shanghai output moves, it moves cars into Canadian driveways too, since Canada is among the export destinations. (Reuters, via Kat Country: https://katcountry989.com/2026/11/03/teslas-china-made-ev-sales-rise-10-3-y-y-in-october/ ; November 2 brief: https://futuretechpulse.ca/china-tech-today-2026-11-02/)

The big picture

Today’s stories share a thread: China’s tech sector is learning to live with constraints — regulatory, hardware and financial. Apple works through a local model to reach Chinese users, DeepSeek reportedly trains on domestic chips, and YMTC waits out a market that no longer funds chip dreams on sight. The breakthroughs have not stopped; the conditions for them have just got harder.

Watch next

The public release of Apple Intelligence in China with iOS 27.3 in December — whether the beta’s Qwen integration survives unchanged, and how Chinese users respond to an iPhone AI that answers through Alibaba.

Sources

1 thought on “China Tech Today: Apple Intelligence Arrives in China — November 4, 2026”

  1. Pingback: China Tech Today: The Self-Reliance Squeeze — Oracle Pulls Tencent's $7 Billion Chip Lease — November 5, 2026 | FutureTechPulse

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top