Alibaba Qwen AI hit by leadership exodus after Qwen3.5 launch

Qwen has been China’s most credible open‑source counterweight to Western frontier labs, and this kind of leadership shock at the very moment of a technically impressive release is strategically significant. The cross‑reports are consistent: Qwen3.5’s small models delivered state‑of‑the‑art “intelligence density” in the 0.8B–9B range, then the architect and several senior contributors abruptly exited, with Alibaba bringing in a former Google DeepMind Gemini leader to take over. That looks like a deliberate pivot from researcher‑led exploration toward a more metrics‑ and monetization‑driven corporate regime.([simplywall.st](https://simplywall.st/stocks/us/retail/nyse-baba/alibaba-group-holding/news/alibaba-qwen-ai-leaders-exit-as-execution-risks-draw-investo?utm_source=openai))

For the race to AGI, the risk isn’t that Qwen disappears—Alibaba has deep pockets and a massive user base—but that one of the most vibrant open‑weight efforts in the world becomes more conservative and closed. If future Qwen flagships go API‑only, we lose an important source of diversity in high‑end model families outside the U.S. ecosystem, and the open‑source community loses a key anchor project. At the same time, a more product‑centric Qwen tightly integrated with Alibaba’s cloud, devices and enterprise stack could accelerate real‑world deployment of agentic systems across Asia at enormous scale.([news.sina.com.cn](https://news.sina.com.cn/zx/ds/2026-03-04/doc-inhpufuy4203509.shtml))

This episode also highlights a broader pattern: as labs move from “heroic research phase” to multi‑billion‑dollar businesses, governance, incentives and talent retention become existential. Qwen’s turmoil rhymes with internal tensions at OpenAI and Google—scientific ambition colliding with commercial pressure. How Alibaba manages this transition will be closely watched by every team trying to keep leading researchers motivated while still hitting growth and safety targets.