Alibaba releases upgraded version Qwen3: fully compatible with Apple's MLX architecture
On June 16th, Alibaba Tongyi Qianwen officially announced the release and open-sourcing of the entire Qwen3 series of 32 MLX quantitative models. MLX is an open-source machine learning framework specifically optimized for Apple chips. The MLX framework can efficiently train and deploy large AI models and is increasingly adopted by AI developers. The official Tongyi Qianwen stated that the company recently updated different quantization versions of the Qwen3 model, including GGUF, AWQ, and GPTQ. It is reported that Alibaba has launched all Qwen3 series models deeply optimized based on the MLX framework, with each model having four different quantization versions: 4bit, 6bit, 8bit, and BF16. In other words, a total of 32 official Qwen3 MLX models are open-sourced at once. Now, from Mac Pro, Mac Studio to Mac mini, MacBook, and even iPad, as well as smaller devices like iPhone, Qwen3 can be easily deployed, truly achieving full-scenario coverage
According to Zhitong Finance APP, on June 16, Alibaba (09988) Tongyi Qianwen officially announced the release and open-sourcing of the entire Qwen3 series of 32 MLX quantitative models. MLX is an open-source machine learning framework specifically optimized for Apple chips, capable of efficiently training and deploying large AI models, and is increasingly adopted by AI developers.
The official Tongyi Qianwen stated that recently, the company updated different quantization versions of the Qwen3 model, including GGUF, AWQ, and GPTQ.
It is reported that Alibaba launched all Qwen3 series models based on the deeply optimized MLX framework, with each model having four different quantization versions: 4bit, 6bit, 8bit, and BF16. In other words, a total of 32 official Qwen3 MLX models are open-sourced at once.
Now, from Mac Pro, Mac Studio to Mac mini, MacBook, and even smaller devices like iPhone, Qwen3 can be easily deployed, truly achieving full-scenario coverage
