This post is also available in:
עברית (Hebrew)
Alibaba Cloud is preparing for the release of its next-generation AI models, Qwen 3, in April. The new series, which includes several different versions, is designed to enhance the company’s growing focus on AI technology, with key improvements aimed at optimizing both cost and efficiency in cloud computing.
According to a report be the South China Morning Post, The Qwen 3 family features a mixture-of-experts (MoE) architecture, a technique used to improve the training process by allocating resources more efficiently. This approach, which was also present in previous versions such as Qwen 1.5 and Qwen 2.5, allows Alibaba to scale its AI models while minimizing costs. One of the smaller variants in the Qwen 3 lineup is equipped with 600 million parameters, making it lightweight enough to be run on mobile devices. This flexibility highlights Alibaba’s push to bring advanced AI capabilities to a wider range of devices and users.
Alibaba’s investment in AI is part of a broader strategy to revolutionize its commerce operations and modernize industries through its cloud services. The company has committed to a substantial $52 billion investment in AI infrastructure over the next three years. This significant capital infusion underscores Alibaba’s ambition to remain at the forefront of the competitive AI landscape, especially as Chinese competitors continue to rise.
Alongside the Qwen 3 release, Alibaba recently launched Qwen Omni, a smaller model designed for the development of cost-effective AI agents. Additionally, the company introduced the QwQ-32B, a 32 billion-parameter model that rivals other leading models in terms of performance, including DeepSeek’s R1 and OpenAI’s o3-mini and o1.
With AI rapidly transforming industries across the globe, Alibaba faces stiff competition. However, Alibaba’s Qwen models continue to push the envelope, with the previous flagship model, Qwen2.5-Max, maintaining a solid position in the global rankings.