Summary

The article discusses the recent resignations of key members from Alibaba's Qwen team, which has developed the promising Qwen 3.5 family of open weight models. The departures raise concerns about the future of Qwen and its innovative AI models.

Key Points

• Junyang Lin, the lead researcher for Qwen, announced his resignation on March 4, 2026, amid a reorganization at Alibaba.
• The resignation triggered a wave of departures from the Qwen team, including Binyuan Hui and Bowen Yu, who were pivotal in model development.
• Qwen 3.5 models were launched starting February 17, 2026, with the largest model being Qwen3.5-397B-A17B at 807GB.
• Smaller models in the family include 122B, 35B, 27B, 9B, 4B, 2B, and 0.8B sizes, with positive feedback on the 27B and 35B models for coding tasks.
• The 2B model is notable for being a multi-modal model that is only 4.57GB in size.
• Alibaba's CEO Wu Yongming addressed the Qwen team in an emergency meeting, indicating the importance of retaining talent.

Analysis

The resignations from the Qwen team signify a potential disruption in the development of innovative AI models, particularly as the Qwen 3.5 series has shown exceptional capabilities. The loss of key personnel could hinder the progress of ongoing projects and affect the competitive landscape in AI development.

Conclusion

IT professionals should closely monitor the developments surrounding Qwen and consider the implications of team dynamics on AI project success. Engaging with emerging models and understanding the shifts in talent within the industry could provide strategic advantages.