Tech giant Alibaba has unveiled Qwen3, the latest addition to its open-source large language model (LLM) family. This marks Alibaba’s debut in hybrid reasoning models, offering advanced AI capabilities to developers worldwide.
The Qwen3 series boasts six dense models and two Mixture-of-Experts (MoE) models, designed to drive the next generation of applications across various platforms—including mobile devices, smart glasses, autonomous vehicles, and robotics. All models are free to download and are available globally.
According to Alibaba, Qwen3 models can seamlessly switch between “thinking mode” for complex, multi-step tasks like mathematics, coding, and logical deduction, and “non-thinking mode” for fast, general-purpose responses. This flexibility allows developers to tailor AI solutions to specific needs.
Since its initial launch, the Qwen model family has garnered over 300 million downloads worldwide, reflecting a growing interest in advanced AI tools. The Chinese mainland has seen a strong demand for LLMs, with various products like DeepSeek, Baidu’s ERNIE, and ByteDance’s Doubao entering the market.
Data suggests that by 2024, nearly 200 generative AI models will have been registered and launched for service in the Chinese mainland, attracting over 600 million registered users. This surge highlights the region’s expanding role in the global AI landscape.
