Alibaba's (BABA, Financial) AI arm just optimized its open-source Qwen3 models for Apple's (AAPL, Financial) MLX framework, a move that likely paves the way for Apple Intelligence to launch in mainland China.
On Monday, Alibaba's Qwen team announced on X that Qwen3 is now available in MLX format with four quantization options—4-bit, 6-bit, 8-bit and BF16—so the models can run locally on iPhones, iPads and Macs.
Released on Hugging Face yesterday, the update aligns with reports in the South China Morning Post that Apple needs a home-grown partner to meet Beijing's tight AI regulations before rolling out its Apple Intelligence suite—features like Genmoji and Writing Tools that elsewhere rely on OpenAI's GPT.
Because any AI models bundled into Apple Intelligence require Chinese regulatory approval, Apple struck a deal with Alibaba to host and vet those models on Chinese soil.
China is Apple's single largest market outside the U.S., and localizing intelligence features could boost iPhone, iPad and Mac sales there—especially as the trade war and data-sovereignty rules complicate direct deployments of Western AI services. Alibaba's update is a clear signal that Apple and its Chinese partner are lining up to meet both user demand and regulatory hurdles.
