Recently, the official Qwen of Alibaba's Tongyi announced the release and open-sourcing of the entire series of 32 MLX quantized models in the Qwen3 series. This move has attracted significant attention in the AI community.

MLX, an open-source machine learning framework specifically optimized for Apple chips, boasts efficient training and deployment capabilities for large AI models. It is gaining increasing favor among AI developers. This action by Alibaba is seen by industry insiders as preparation for Apple Intelligence (Apple Intelligence) in mainland China. Previously, authoritative sources have revealed that Alibaba will become Apple’s large-scale model partner in mainland China.

Preparation for Apple Intelligence in Mainland China! Alibaba Releases Upgraded Qwen3: Fully Compatible with Apple MLX Architecture

It is understood that the Qwen3 series models launched this time by Alibaba have undergone deep optimization based on the MLX framework. Each model in this series provides four different precision quantization versions: 4-bit, 6-bit, 8-bit, and BF16, meaning a total of 32 official Qwen3MLX models have been fully open-sourced at once.

The Qwen3 series models released this time are highly adaptable. From Mac Pro, Mac Studio to Mac mini, MacBook, iPad, and even iPhones with smaller memory, they can be easily deployed, truly achieving full-scene coverage, and bringing more possibilities to users of Apple devices.