CN — TechNode

Apple’s China AI partner has had a rough ride

AINovember 6, 20261 min read

Alibaba's Qwen AI models are now part of Apple Intelligence tests in China. The story looks at how Qwen grew from an uncertain project inside Alibaba into a major global AI model family.

Qwen Joins Apple's China Launch

On November 4, Apple Intelligence began beta testing in China. Alibaba's Qwen is available in some Siri, Writing Tools and Visual Intelligence features. The partnership marks a fast rise for a model family that was first launched in 2023.

Key Facts

Alibaba introduced Tongyi Qianwen, the first version of Qwen, in April 2023.
Hugging Face counted more than 2 billion Qwen downloads on its own Hub in the first seven months of 2026, well ahead of Google and Meta.
Qwen3, released in April 2025, came with eight open models.
Qwen3's largest open model activates 22 billion of its 235 billion parameters at a time to lower the cost of running it.

A Hard Road to the Top

Qwen did not win inside Alibaba right away. Some teams kept using Meta's Llama into 2024. Over time, open source became central to Qwen's strategy. Developers improved the models and sent feedback back to the team. DeepSeek's R1 model then raised the pressure across China's AI industry, pushing rivals to offer strong reasoning at lower cost.

Trouble Inside the Team

In early 2026, the problems were not only technical. Technical leader Junyang Lin left in March, and other senior researchers also departed. Reports said the disputes involved reorganization, hiring and computing resources. In response, Alibaba grouped key model teams under a new unit called Token Foundry and named Dayiheng Liu to lead the Qwen large language model project.

Why It Matters

Qwen's story shows how hard it is to build top AI models. These models need huge amounts of money and computing power. Strong research teams, however, often need freedom to move fast. How Alibaba handles this balance may affect whether Qwen stays a leader in the global AI race.

Glossary

Qwen: Alibaba's family of AI language models, which can write, answer questions and reason.
Apple Intelligence: Apple's set of AI features built into its devices, such as Siri and Writing Tools.
Hugging Face: A company whose website hosts AI models and tools that developers can download.
open source: Software whose code or model files are shared publicly so anyone can use, change and improve them.
DeepSeek: A Chinese AI company whose R1 reasoning model pushed rivals to improve their own models.

Fuji TV to Use Generative AI in All Dramas by 2027; Up to 80% Labor Cut in Some Production Steps

Fuji Television announced on November 5 that it will have a system in place by October 2027 to use generative AI in all dramas it produces. Starting with the October 2026 season, at least one drama per quarter will be designated an "AI Strengthened Slot," and the company says some video production steps have already cut labor by up to 70–80% compared with conventional methods, though this applies only to specific steps, not whole-production costs.

AIThe VergeNov 6

OpenAI is stepping back from NDAs for data center projects

After this week's midterms, in which voters voiced opposition to data center expansion, OpenAI announced new commitments to communities where it works on data center projects. These include not using non-disclosure agreements with public officials, paying the full cost of the electricity and infrastructure it needs, and publishing expected jobs and hiring reports. OpenAI also said future "compute campuses" it owns and operates will have a "local compact" reflecting local priorities.

AI전자신문Nov 6

OpenAI Releases 'GPT-6 Cyber,' a Cybersecurity-Focused AI Model

OpenAI released GPT-6 Cyber, an AI model specialized in cybersecurity, on the 6th. It is first available to approved "Daybreak Red" customers via API and Codex, with plans to expand to all Daybreak Red customers. In OpenAI's tests, the model found 81 high-risk vulnerabilities in Firefox and reached a 48.5% success rate on the ExploitGym benchmark, up from 32.1% for its predecessor, GPT-5.6 Cyber.

AIITmediaNov 6

OpenAI Unveils Cybersecurity-Focused "GPT-6 Cyber," Finds 81 Unknown Firefox Vulnerabilities

OpenAI announced GPT-6 Cyber, a cybersecurity-specialized AI model built on the general-purpose GPT-6 Astra, and will initially offer it only to approved users of its Daybreak Red defender program. In research on real software, the model found 81 unknown high-severity vulnerabilities in Firefox and developed working exploits for 948 Linux kernel vulnerabilities, and OpenAI says it is also restricting access for critical infrastructure organizations.