Meta Announces Official Launch of 'Superintelligence Laboratory' with 11 Top Talents Revealed}

Meta has made a new move! Recently, Meta has been aggressively recruiting top AI talent, having already poached around ten employees from OpenAI, with seven publicly confirmed so far.

Today, according to Bloomberg and other sources, Mark Zuckerberg announced in an internal memo on Monday that the company's AI research, infrastructure, and product teams will be reorganized into the newly established "Meta Superintelligence Labs" (Meta SML).

This new department will be led by Alexandr Wang, former CEO of Scale AI, who will serve as Meta’s Chief AI Officer. Zuckerberg also revealed for the first time 11 top talents recruited from OpenAI, Anthropic, Google DeepMind, and other leading institutions.

Zuckerberg stated that MSL will integrate various teams to focus on developing the open-source Llama series large models, related products, and foundational AI research projects.

Below is the full content of Zuckerberg’s memo:

With the rapid advancement of AI, the development of superintelligence is imminent. I believe this marks the beginning of a new era for humanity, and I am committed to ensuring Meta leads this process. Today, I want to detail how we are restructuring our organization to realize our vision: "Building personalized superintelligence for everyone."

We will rename the entire organization as "Meta Superintelligence Labs" (Meta SML). This includes all our basic research, product teams, FAIR teams, and a newly established lab dedicated to next-generation models.

Alexandr Wang has joined Meta as our Chief AI Officer (Chief AI Officer) and will lead MSL. Having worked with me for many years, I consider him one of the most outstanding peer entrepreneurs. He has a clear understanding of the historic significance of superintelligence and, as co-founder and CEO, has helped ScaleAI grow rapidly, participating in the development of nearly all leading industry models.

Nat Friedman has also joined Meta to co-lead MSL alongside Alexandr, responsible for our AI products and application research. Nat will work with Connor to define his future responsibilities. He previously managed Microsoft’s GitHub and now runs a leading AI investment firm. Over the past year, Nat has served on Meta’s advisory board, gaining deep insight into our roadmap and strategic needs.

Additionally, several new strong members have joined recently or in recent weeks, and I am pleased to announce:

• Trapit Bansal — Pioneer in Chain-of-Thought (CoT) reinforcement learning, co-creator of OpenAI’s series models.
• Shuchao Bi — Co-creator of GPT-4o voice mode and o4-mini, former head of multimodal training at OpenAI.
• Huiwen Chang — Co-creator of GPT-4o image generation system, inventor of MaskGIT and Muse architectures at Google Research.
• Ji Lin — Developer of o3/o4-mini, GPT-4o, GPT-4.1, GPT-4.5, GPT-4o-imagegen, and Operator inference framework.
• Joel Pobar — Expert in reasoning systems at Anthropic, previously led HHVM/Hack/Flow/Redex development and performance tools at Meta for 11 years.
• Jack Rae — Head of Gemini pretraining and architect of Gemini 2.5 inference framework, led early large model development at DeepMind (Gopher/Chinchilla).
• Hongyu Ren — Co-creator of GPT-4o/4o-mini/o1-mini/o3-mini/o3/o4-mini, former head of post-training at OpenAI.
• Johan Schalkwyk — Former Google Fellow, early contributor to Sesame system, technical lead of Maya project.
• Pei Sun — Architect of DeepMind Gemini post-training, programming, and reasoning models, led perception models at Waymo.
• Jiahui Yu — Co-creator of o3/o4-mini, GPT-4.1, GPT-4o, former head of perception team at OpenAI, co-leader of Gemini multimodal system.
• Shengjia Zhao — Co-creator of ChatGPT, GPT-4/4.1/o3, former head of synthetic data team at OpenAI.

I am excited about the progress of Llama 4.1 and 4.2 models. These models power Meta AI, which now has over 1 billion monthly active users across our applications, with an ever-growing intelligent network continuously optimizing our products and technology. We will relentlessly push these models’ iterations and upgrades.

Meanwhile, our next-generation model research will aim to reach the technological peak within about a year. Over the past months, I have visited Meta’s internal teams, top AI labs, and promising startups to assemble the core team for this elite project. The team is still expanding, with more top talents to be invited from within the company’s AI ecosystem.

Meta holds a unique advantage in delivering superintelligence: we have strong business support, massive computing resources beyond small labs, extensive experience in product development and growth for billions of users, and are pioneering rapidly expanding fields like AI glasses and wearables. Our corporate structure also empowers us to advance strategies with confidence and boldness.

I firmly believe that this dual strategy of talent influx and model development will lay a solid foundation for "realizing personalized superintelligence for all." More top talents will join various teams in the coming weeks. Stay tuned. I am ready to dedicate myself fully to this mission.

Currently, new Meta recruits are already promoting on social media, such as Jack Rae.

^{Reference Links:}

^{https://www.bloomberg.com/news/articles/2025-06-30/zuckerberg-announces-meta-superintelligence-effort-more-hires?srnd=phx-technology}

^{https://www.cnbc.com/2025/06/30/mark-zuckerberg-creating-meta-superintelligence-labs-read-the-memo.html}

^{https://www.theverge.com/news/695355/mark-zuckerberg-meta-ai-superintelligence-labs}