Alibaba to release its flagship Qwen models this month

Several Qwen 3 models are to be released this April, according to the media.

Cloud computing company Alibaba Cloud, a subsidiary of the Chinese conglomerate Alibaba Group, is preparing to launch its Qwen 3 family of models, claim Reuters and Alibaba-owned South China Morning Post, citing sources.

According to the South China Morning Post, Qwen 3 features a mixture-of-experts (MoM) architecture, which allows for cost-efficient training. Previous company models, including Qwen 1.5 and Qwen 2.5, also featured MoM.

One of the smaller Qwen 3 models is said to come in 600 million parameters and can be run on mobile devices.

Alibaba is betting heavily on AI to transform its commerce operations and upgrade traditional industries through its cloud business. Over the next three years, the company plans to invest $52 billion in building out AI infrastructure.

Last week, Alibaba released Qwen Omni, a smaller model focused on developing cost-effective AI agents.

Recently, the company also unveiled a smaller 32 billion parameter model, QwQ-32B, claiming comparable performance to DeepSeek’s R1 and OpenAI’s o3-mini and o1.

Alibaba is one of the many companies competing in the fast-changing field of AI development. Last week, Google released Gemini Pro 2.5 Experimental, which claims the top spot in the crowdsourced AI benchmarking platform Lmarena.

The previous flagship, Alibaba’s Qwen2.5-Max, is currently ranked 11th.