Qwen shifts to closed source after firing lead researcher Junyang Lin
Alibaba’s Qwen abandons open source, leaving DeepSeek and others as sole open-source leaders.
Qwen, Alibaba’s open-source AI lab, has abruptly stopped releasing large open-source models after firing Junyang Lin, the lead researcher and public face of the project. According to Chinese social media reports, the small-model team has also been disbanded. Qwen 3.6 and possibly 3.7, both of which Lin worked on before his departure, are expected to be the last open-source versions from the lab. All subsequent large models will remain fully closed source, marking a dramatic shift in strategy for one of China’s most prominent AI contributors.
This move puts Qwen in stark contrast to its domestic competitors. As of mid-2026, every other major Chinese AI lab has released an open-source model more recently than Qwen. GLM released GLM-5.2 in June 2026, Kimi released Kimi-K2.7-Code on June 12, MiniMax released MiniMax-M3 on June 11, Step released Step-3.7-Flash on June 11, and DeepSeek released DeepSeek-V4-Pro/V4-Flash in April 2026. Qwen now stands alone as the last major Chinese lab without a recent open-source release, raising concerns about the health of China’s open-source AI ecosystem.
- Junyang Lin fired; Qwen's large models now fully closed source.
- Small-model team reportedly gone; Qwen 3.6/3.7 were his last open-source releases.
- Other Chinese labs (GLM, Kimi, MiniMax, Step, DeepSeek) all released open-source models more recently than Qwen.
Why It Matters
Open-source AI competition narrows as Qwen retreats, giving DeepSeek and others a strategic advantage.