On August 3, 2023, Alibaba Cloud released Qwen-7B and Qwen-7B-Chat, the first openly downloadable models in its Qwen (Tongyi Qianwen) family, publishing the weights on Hugging Face and ModelScope along with a technical report. The models were trained on about 3 trillion tokens of multilingual data with a focus on Chinese and English.
Over the following months Alibaba expanded the line into base sizes of 1.8B, 7B, 14B, and 72B parameters, with the Qwen-72B model - supporting a 32K-token context - arriving on November 30, 2023, plus multimodal Qwen-VL variants. The larger models were released under Alibaba’s own Tongyi Qianwen license, which permits broad use with restrictions on the very largest commercial deployments.
Qwen grew into one of the most widely adopted open-weight model families in the world, heavily used as a base for fine-tuning and derivative models. Together with DeepSeek, it established Chinese labs as major contributors to the global open-weights ecosystem rather than just consumers of Western models.