Alibaba Launches and Open-Sources Qwen3, China’s First Hybrid Reasoning AI Model
On April 29, Alibaba officially unveiled Qwen3, marking a significant milestone as China’s first hybrid reasoning large language model. This new series integrates both fast and slow thinking modes, allowing users to toggle between rapid responses and in-depth, step-by-step reasoning based on task complexity. This flexible design aims to balance computational efficiency with high-level intelligence. The Qwen3 series includes several models, notably the flagship Qwen3-235B-A22B, a Mixture-of-Experts (MoE) model with over 235 billion parameters, and the lightweight Qwen3-30B-A3B, which features 30 billion total parameters with only 3 billion active during inference. Alibaba Cloud has open-sourced these models, making them available across major platforms. According to the company, Qwen3-235B-A22B delivers competitive performance in coding, mathematics, and general reasoning benchmarks, rivaling top global models such as DeepSeek-R1, Grok-3, and Gemini 2.5 Pro. This launch underscores Alibaba's continued commitment to advancing artificial intelligence capabilities while promoting open-source accessibility in the tech community.
Wire timeline
Alibaba Launches and Open-Sources Qwen3, China’s First Hybrid Reasoning AI Model
On April 29, Alibaba officially unveiled Qwen3, marking a significant milestone as China’s first hybrid reasoning large language model. This new series integrates both fast and slow thinking modes, allowing users to toggle between rapid responses and in-depth, step-by-step reasoning based on task complexity. This flexible design aims to balance computational efficiency with high-level intelligence. The Qwen3 series includes several models, notably the flagship Qwen3-235B-A22B, a Mixture-of-Experts (MoE) model with over 235 billion parameters, and the lightweight Qwen3-30B-A3B, which features 30 billion total parameters with only 3 billion active during inference. Alibaba Cloud has open-sourced these models, making them available across major platforms. According to the company, Qwen3-235B-A22B delivers competitive performance in coding, mathematics, and general reasoning benchmarks, rivaling top global models such as DeepSeek-R1, Grok-3, and Gemini 2.5 Pro. This launch underscores Alibaba's continued commitment to advancing artificial intelligence capabilities while promoting open-source accessibility in the tech community.
TechNode