Alibaba Qwen Leader: The Agent Era Has Arrived

At the 2026 Apsara Conference held in Hangzhou on September 22, 2026, Liu Dayiheng, Head of the Qwen LLM project at Alibaba's ATH business group, delivered a speech on the latest trends in large language models and the progress of the Qwen series.

Liu said that in 2026, large models have officially entered the third stage: the agent era. Models are no longer assistants that passively wait for questions; they can autonomously plan, call tools, iterate in closed loops, and independently complete tasks. The consumption pattern of AI has shifted from single-turn conversations to complete tasks, driving an exponential surge in global demand for tokens.

He also noted that over the past few years, Qwen has continuously open-sourced advanced models across modalities and sizes. To date, the Qwen series has surpassed 3 billion global downloads, with more than 300,000 derivative models, ranking first worldwide. The team has open-sourced more than 460 models in total. Recently, Qwen 3.8 27B, released just two months ago, became the most-liked model on HuggingFace, surpassing well-known models such as DeepSeek and Llama.

Liu emphasized that AI value should not belong only to a few enterprises, but should become a foundational capability accessible, usable, and creatable by users worldwide. In the future, the team will continue to uphold openness and open source, enabling more people to afford, use, and participate in creating AI, achieving AI for all.