2025年10月23日，Meta宣布對其人工智能部門進行大規模裁員，涉及約600名員工，範圍涵蓋FAIR（基礎AI研究團隊）、AI基礎設施及產品團隊。此次裁員引發業界震動，其中最受關注的是資深研究員田渊栋及其團隊被整體裁撤。

裁員由Meta首席AI官Alexandr Wang主導，旨在精簡組織結構、提升運營效率。被裁員工主要集中於非核心團隊，新成立的TBD Lab未受影響。據了解，田渊栋團隊在Llama 4項目後期承擔大量調試工作後被裁，他公開表示團隊“完成髒活累活後遭卸磨殺驢”，事件也暴露出Meta內部存在資源爭奪和架構混亂的問題。

被裁員工將獲得16周的基礎遣散費，每工作滿一年額外補償兩周薪資（扣除通知期）。田渊栋透露，個人獲得8個月薪資的賠償。部分員工的離職日期定於11月21日，過渡期內可嘗試內部轉崗，但部分權限已被關閉，實際轉崗成功率尚不明朗。

此裁員事件迅速在行業內引發反應，OpenAI、xAI、英偉達、Anthropic等領先企業及初創公司紛紛通過社交媒體向被裁員工伸出橄欖枝。田渊栋在推特上的發文瀏覽量超過300萬，評論區成為人才招募現場。業界人士如姚順雨、Oriol Vinyals公開招攬人才，Meta因此被指損失大量頂尖研究者，人才市場競爭異常激烈。

事件同時反映出Meta內部矛盾，FAIR與產品團隊長期存在資源競爭。Alexandr Wang通過裁員鞏固地位，使Meta策略逐步向閉源模型和商業化傾斜，學術研究團隊的生存空間被擠壓。此外，楊立昆（Yann LeCun）澄清未參與Llama 4項目，Meta被曝因Llama 4.5訓練完成而裁員，被批評“過河拆橋”。

裁員消息發酵後次日，Meta宣布將裁撤風控部門，改用AI自動化合規流程，進一步引發外界爭議。分析人士指出，Meta未來可能繼續裁員，重心將向“超級智能”研發傾斜，但大規模人才流失可能對其長期競爭力造成影響，行業也將密切關注其後續策略調整。

On October 23, 2025, Meta announced a major layoff affecting approximately 600 employees in its artificial intelligence division, including teams from FAIR (Fundamental AI Research), AI infrastructure, and product development. The move sent shockwaves through the tech industry, particularly as senior researcher Tian Yuandong and his entire team were laid off.

The layoffs were led by Meta’s Chief AI Officer Alexandr Wang and aimed at streamlining organizational structure and improving operational efficiency. The affected employees were primarily from non-core teams, while the newly established TBD Lab remained unaffected. Tian Yuandong revealed that his team, which had handled extensive fine-tuning work on the Llama 4 project, was let go after completing the “dirty and laborious tasks,” highlighting internal resource conflicts and structural disorganization within Meta.

Compensation for laid-off employees includes a 16-week base severance package, with an additional two weeks’ salary per year of service (excluding the notice period). Tian disclosed that he personally received eight months’ salary in severance. Some employees are set to leave on November 21, with a transitional period allowing for internal transfers, though certain permissions have been restricted, and the actual success rate for transfers remains uncertain.

The layoffs triggered swift industry reactions, with leading companies such as OpenAI, xAI, NVIDIA, and Anthropic, along with numerous startups, reaching out to affected employees via social media. Tian’s tweet announcing the layoffs garnered over 3 million views, with the comment section becoming a de facto recruitment platform. Industry figures like Yao Shunyu and Oriol Vinyals publicly courted talent, highlighting Meta’s loss of top researchers and intensifying competition for AI professionals.

The event also underscored internal tensions at Meta, as FAIR and product teams have long competed for resources. Wang’s move to consolidate power reflects a strategic shift toward closed-source models and commercialization, squeezing the space for academic research within the company. Additionally, Yann LeCun clarified that he was not involved in the Llama 4 project, and reports suggest layoffs were tied to the completion of Llama 4.5 training, drawing criticism for “cutting off those who helped build the project.”

The day after the layoffs, Meta announced plans to dismantle its risk control department, replacing it with AI-driven automated compliance processes, sparking further controversy. Analysts suggest that Meta may continue to pursue layoffs while focusing on “superintelligent” AI development, but the large-scale loss of talent could negatively affect its long-term competitiveness. The industry is closely watching how Meta navigates this strategic transition.