AI

Junyang Lin Steps Down as Qwen Tech Lead in Abrupt Departure

On March 3, 2026, Junyang Lin — the tech lead and public face of Alibaba’s Qwen open-source AI project — announced his departure in a brief post on X: “me stepping down. bye my beloved qwen.” The exit came just 24 hours after the Qwen team shipped the Qwen 3.5 small model series. At least two other Qwen researchers left around the same time. Alibaba has not commented on the reasons for any of the departures.

Who Is Junyang Lin?

Lin joined Alibaba in 2019 as a Senior Algorithm Engineer working on NLP and multimodal research. By 2023, he had become the formal tech lead of the Qwen team, steering the project from a nascent lab effort into a global open-source powerhouse. Under his leadership, the Qwen family expanded from early 7-billion-parameter language models into a sprawling portfolio spanning vision-language models (Qwen-VL), audio models, math-specialized models, coding agents, and the reasoning-focused QwQ series.

The numbers speak for themselves: over 600 million downloads and more than 170,000 derivative models on Hugging Face. On Google Scholar, Lin has accumulated over 42,000 citations, with the Qwen3 technical report alone drawing nearly 9,000. Before Qwen, he played central roles in Alibaba’s extreme-scale MoE model M6, OFA (published at ICML 2022, 700+ citations), and Chinese-CLIP. He was also a core maintainer of OpenDevin.

Beyond the technical contributions, Lin was Qwen’s chief communicator — regularly announcing model releases, sharing benchmark results, and engaging directly with the global developer community building on top of Qwen’s models.

What Happened

Lin’s one-line announcement gave no explanation. Chen Chang, a Qwen contributor, wrote: “I’m truly heartbroken. I know leaving wasn’t your choice. Just last night, we were side by side launching the Qwen3.5 small model. I honestly can’t imagine Qwen without you.”

Lin was not alone. At least two other researchers exited around the same time:

  • Kaixin Li announced his departure, writing: “Signing off from @Alibaba_Qwen. Grateful for the chance to work with such brilliant minds.”
  • Binyuan Hui updated his X profile to read “former MTS at Qwen,” indicating his exit as well.

Alibaba has not issued any public statement about the departures or the Qwen team’s leadership structure going forward.

What This Means for Qwen

The timing is striking. Qwen 3.5 has been one of the most impactful open-source AI releases of 2026 — the small models alone demonstrated that a 9B-parameter model could outperform models 3–13× its size across language, vision, and agentic benchmarks. The project was on a clear upward trajectory.

Losing Lin removes the visible human element that distinguished Qwen in international discourse. While Alibaba maintains operational continuity — the models are released, the infrastructure exists, and the broader team remains — the departure of the tech lead, the community ambassador, and at least two other researchers in a single week represents a significant loss of expertise for Alibaba’s AI division.

For the open-source AI community, which has increasingly relied on Qwen models as high-quality alternatives to closed-source systems, the key question is whether development momentum and the team’s open-source ethos will survive this leadership transition. Wenting Zhao, a research scientist on the Qwen team, described Lin’s departure simply as “the end of an era.”

This post was drafted with AI assistance and reviewed by RITS staff.

Sources

Related stories

AI

Qwen 3.5 Small Models: 9B Parameters That Beat 120B

Key points On March 2, 2026, Alibaba’s Qwen team completed the Qwen 3.5 family with four small dense models of 0.8B, 2B, 4B, and 9B parameters for on-device and edge deployment. All four models are open-weight under Apache 2.0 and available on Hugging Face and ModelScope. Qwen3.5-9B outperforms models 3-13 times its size across language, […]

agentic ai

Qwen 3.5 Medium Series: Frontier AI That Fits on Your GPU

Key points On February 24, 2026, Alibaba’s Qwen team expanded the Qwen 3.5 family with three Apache 2.0 open-weight models — Qwen3.5-27B, Qwen3.5-35B-A3B, and Qwen3.5-122B-A10B — available on Hugging Face. Qwen3.5-27B is a dense model with all 27 billion parameters active at inference, optimized for coding and instruction-following, and uses a 3:1 hybrid ratio of […]

agentic ai

Qwen 3.5: Alibaba’s Native Multimodal Agent Model Arrives

Key points Alibaba’s Qwen team released Qwen 3.5 on February 16, 2026, with a flagship 397B-parameter mixture-of-experts model built for the agentic AI era. The flagship Qwen3.5-397B-A17B activates only 17 billion parameters per forward pass despite having 397 billion total parameters. Listed specifications include a 256K-token native context window, 1M tokens on the hosted Qwen3.5-Plus, […]

agentic ai

Qwen3-Coder-Next: Alibaba’s Ultra-Sparse 80B Coding Agent

Key points Alibaba’s Qwen team released Qwen3-Coder-Next in February 2026 under the Apache 2.0 license, an open-weight coding model that activates only 3 billion of its 80 billion total parameters per forward pass. On SWE-Bench Verified with the SWE-Agent scaffold the model scores 70.6%, and it also scores 62.8% on SWE-Bench Multilingual and 44.3% on […]

Never miss a RITS update

Subscribe to the RSS feed for AI, emerging technology, and digital scholarship news curated for the NYU Shanghai community, or get in touch to suggest a story.