Chinese technology giant Alibaba Group has unveiled a new generation of its leading artificial intelligence models, Qwen3, reaffirming its determination to remain at the forefront of the global AI race, which has accelerated since the beginning of 2025 following the emergence of a powerful model from DeepSeek.
The Qwen3 series includes several advanced models, among which are two based on ‘mixture-of-experts’ (MoE) technology, an increasingly popular approach that divides tasks among multiple specialized sub-models, similar to how an expert team addresses complex problems. Such systems enable faster, more accurate, and energy-efficient execution of tasks such as mathematical calculations and programming. In these areas, Alibaba claims that Qwen3 competes with and even surpasses DeepSeek.
The company specifically highlighted that the new generation of models significantly reduces implementation costs compared to competitors, aiming to attract businesses and developers seeking powerful yet accessible AI tools.
Everything in the Name of Qwen
Since CEO Eddie Wu declared in February that the company’s primary goal is the development of artificial general intelligence (AGI), Alibaba has accelerated the pace of launching new AI solutions. Just a few weeks ago, they introduced the Qwen 2.5 model, capable of processing text, images, sound, and video – with sufficiently optimized performance to be used on mobile phones.
In March, a new version of the Quark application was launched, a digital assistant that utilizes Alibaba’s AI models for everyday tasks such as searching, translating, and managing documents. All Qwen3 models will be open source, representing a significant step towards more transparent development of artificial intelligence, especially in the context of increasing demands for regulatory oversight of advanced AI systems.
