Organizational Instability And Attrition Signals

Issue 63 Edition 2026-03-04 6 min read

General

Sources: 1 • Confidence: High • Updated: 2026-03-08 21:22

Key takeaways

Junyang Lin announced on X that he is stepping down from the Qwen team.
A stated watch item is whether departing core Qwen members form a new effort or join another research lab, potentially shifting open-model innovation to a new organization.
The Qwen3.5-397B-A17B release is described as an 807GB model and was followed by 122B, 35B, 27B, 9B, 4B, 2B, and 0.8B variants.
A possible trigger for Lin’s resignation was an Alibaba re-org that put a newly hired researcher from Google’s Gemini team in charge of Qwen.
Reporting summarized from 36Kr states that, as of the emergency meeting, no conclusion was reached on Lin Junyang’s whereabouts or whether he would return, despite a later WeChat Moments post encouraging Qwen colleagues to continue as planned.

Junyang Lin announced on X that he is stepping down from the Qwen team.
Reporting summarized from 36Kr states that, as of the emergency meeting, no conclusion was reached on Lin Junyang’s whereabouts or whether he would return, despite a later WeChat Moments post encouraging Qwen colleagues to continue as planned.
Reporting summarized from 36Kr states that Tongyi Lab held an emergency all-hands meeting around 1:00 PM Beijing time on March 4, attended by Alibaba Group CEO Wu Yongming, in response to the situation.
Reporting states that several other key Qwen members resigned the same day, including leaders responsible for Qwen code development, post-training, and contributions to Qwen 3.5/VL/Coder.
Reporting summarized from 36Kr states that Lin Junyang announced his resignation at 0:11 AM Beijing time on March 4 and that the departure caused significant internal and industry reaction.

A stated watch item is whether departing core Qwen members form a new effort or join another research lab, potentially shifting open-model innovation to a new organization.
Alibaba CEO attendance at the emergency all-hands has been interpreted as suggesting Alibaba may try to retain some departing Qwen talent, while the situation remains unsettled.
It is asserted that if the Qwen team disbands now, it would significantly undermine continued progress in achieving high-quality results from smaller and smaller models.

The Qwen3.5-397B-A17B release is described as an 807GB model and was followed by 122B, 35B, 27B, 9B, 4B, 2B, and 0.8B variants.
There are positive reports that the Qwen 27B and 35B models perform well for coding tasks while still fitting on 32GB or 64GB Macs.

A possible trigger for Lin’s resignation was an Alibaba re-org that put a newly hired researcher from Google’s Gemini team in charge of Qwen.

A stated watch item is whether departing core Qwen members form a new effort or join another research lab, potentially shifting open-model innovation to a new organization.

What is Junyang Lin’s definitive employment status and role outcome (departure finalized vs return vs reassignment)?
Did an Alibaba re-org place a new leader over Qwen, and if so, who and with what mandate?
How many additional Qwen core members actually resigned, which functions were affected, and were any resignations rescinded?
What concrete actions (if any) did Alibaba/Tongyi Lab take after the emergency all-hands (leadership appointment, retention packages, reprioritization)?
Are the reported coding performance and Mac fit claims for Qwen 27B/35B reproducible under specified inference settings (quantization level, context length, runtime, memory footprint)?

Leadership churn and unresolved departures could slow Qwen release cadence, reduce coordination, and weaken continuity of open model innovation tied to Alibaba and Tongyi Lab.
Departing core Qwen members could form a new effort or join another lab, potentially shifting talent and momentum in open models away from Alibaba to a different organization.
If smaller Qwen3.5 variants are practically deployable on consumer Macs as reported, downstream developer adoption could concentrate on 27B and 35B even if flagship scale models are operationally heavy.

Official clarification of Junyang Lin status and reporting line, plus named leadership for Qwen and any stated mandate following the reported emergency meeting.
Observable talent motion such as additional core resignations, team departures, or public affiliations indicating members joined another lab or launched a new project.
Continued, timely releases and maintenance of Qwen3.5 variants, including reproducible deployment details for 27B and 35B such as quantization, context length, runtime, and memory footprint.

Confirmed retention outcome with Lin returning or remaining in a defined role, and no further verified core departures, reducing the likelihood of structural instability.
No evidence of a spinout or external team formation and no confirmed joins to other labs by departing members over a reasonable monitoring window.
If reported Mac deployability and coding performance claims for 27B and 35B are not reproducible under specified inference settings, weakening the adoption read through.