Alibaba Qianwen, a core figure, declares: Goodbye, Qianwen!

On March 4th, Alibaba senior algorithm expert and head of the Qianwen large model team, Lin Junyang, announced on social media that he will be leaving the Qianwen team. It is understood that Lin Junyang was the leader of Alibaba’s Qianwen large model.

Previously, the Qianwen team launched the lightweight Qwen3.5 model series, with four versions: 0.8B, 2B, 4B, and 9B, all open source and commercially available.

On the evening of March 2nd, Alibaba open-sourced four small-sized Qwen3.5 models: Qwen3.5-0.8B, Qwen3.5-2B, Qwen3.5-4B, and Qwen3.5-9B, which can be deployed on laptops and other edge devices. The release quickly attracted attention from the overseas tech community, with Elon Musk commenting under a post on the Qianwen social platform, “Amazing intelligence density.”

However, Lin Junyang did not explicitly state that he would leave Alibaba, and his next move remains unknown.

Looking at his career history, Lin Junyang is the youngest P10 in Alibaba’s history. He joined Alibaba in 2019 and has been there for nearly 7 years. Before his departure, he served as the head of Alibaba’s Qianwen large model team and was a core member of Alibaba Group’s Tongyi Laboratory.

Born in 1993, Lin Junyang holds a bachelor’s degree in computer science and a master’s in linguistics from Peking University. After graduating with his master’s, he joined Alibaba DAMO Academy’s Intelligent Computing Laboratory in 2019 as an algorithm expert, participating in the development of Alibaba’s M6 multimodal pre-training model.

In 2022, multiple research teams at Alibaba DAMO Academy reorganized within Alibaba Cloud to establish the Tongyi Laboratory. Lin Junyang transferred from DAMO Academy to the Tongyi Laboratory and was officially appointed as the technical lead for the Qianwen series large models.

The Qianwen large models have gradually grown to become the world’s strongest open-source models in the AI large model competition. On April 29, 2025, Alibaba open-sourced the Qwen3 series models, which surpassed top global open-source models in multiple authoritative evaluations, officially ranking as the world’s strongest open-source models. By January 2026, the latest data from the world’s largest AI open-source community, Hugging Face, showed that Alibaba’s Qianwen derivative models exceeded 200,000, with over 1 billion downloads, firmly maintaining the top position among open-source large models worldwide.