Alibaba introduces open-source AI model with hybrid reasoning

HANGZHOU -- China's tech giant Alibaba on Tuesday launched Qwen3, the latest generation of its open-sourced large language model (LLM) family, marking the company's debut of hybrid reasoning models.

The Qwen3 series features six dense models and two Mixture-of-Experts (MoE) models, offering the flexibility required to build next-generation applications across mobile devices, smart glasses, autonomous vehicles, robotics and beyond.

All the models are open-sourced and available worldwide.

Qwen3 models can switch between thinking mode for complex, multi-step tasks such as mathematics, coding, and logical deduction and non-thinking mode for fast, general-purpose responses, according to Alibaba.

Since its debut, the Qwen model family has seen over 300 million downloads worldwide.

China has a strong demand for LLMs with various products, such as DeepSeek, Baidu's ERNIE and ByteDance's Doubao, developed. Data indicates that by 2024, nearly 200 generative AI models had been registered and launched for service in China, with over 600 million registered users.