Alibaba Cloud has rapidly scaled its Qwen artificial intelligence family from a 2023 chatbot release into an open-weight ecosystem featuring models with up to 2.4 trillion parameters. Beginning with the April 2023 introduction of Tongyi Qianwen by then-CEO Daniel Zhang, Alibaba quickly expanded the lineup by open-sourcing the Qwen-7B model in August 2023 and following up with vision capabilities, larger 72B variants, and specialized audio and math versions throughout 2024.
The release tempo accelerated with the Qwen2.5 rollout in late 2024, which introduced over 100 open-source models trained on 18 trillion tokens, alongside coding variants and early reasoning previews. By early 2025, Alibaba countered competitor releases by shipping Qwen2.5-VL with device control, the large-scale MoE Qwen2.5-Max, and the reinforcement learning-trained QwQ-32B model, which achieved performance comparable to much larger systems while driving a stock price increase of over 7 percent.
This rapid multi-year iteration cycle highlights how open-weight strategies and rapid reinforcement learning deployments are reshaping the global artificial intelligence landscape.