Mark Zuckerberg Create Meta Meverintelligence lab. Read the memo

Mark Zuckerberg said on Monday what creates Meta Labs Superintelligence, which will be headed by some of the latest hiring his company, including the scale of AI Ex-Ces Alexandr Wang and former Github NAT CEO Friedman.

Zuckerberg said the new AI Superintelligenc unit projectsAccording to the internal memo obtained by CNBC.

Bloomberg first report about the new block.

Co -founder and Meta CEO were on Oh hire a blitz As he faces fierce competition from competitors such as Openai and Google. Earlier in June, the company said it would be hire Van, who was now the chief officer of the II Meta, and some of his colleagues as part of an investment in $ 14.3 billion.

Meta also hired Friedman and his business partner Daniel Gros, who was the CEO of a safe supertell, a startup AI, created by co -founder Openai Ilya Sutskever, CNBC report. Meta tried to buy a safe superstar, but there was an eahydler from SutsKever.

CEO Openai Sam Altman – Note In a recent podcast that Meta was recruited AI researchers From his company, offering the signing of bonuses up to $ 100 million.

In an interview on June 20, Meta Technology Chief Andrew Bosuort told CNBC “Final Time Bell” that Openai is opposed by Meta’s proposal.

“The market sets a bet here for the talent level, which is really incredible and some unprecedented in my 20-year career as a head of technology,” Boswort said.

Here’s a complete inner memo of Zuckerberg, published on Monday:

As the AI ​​progress is accelerated, the development of the warden is in view. I believe it will be the beginning of a new era for humanity, and I fully try to do what you need to make the way. Today I want to share some details about how we organize our AI efforts to build our vision: personal visibility for everyone.

We are going to name our common organization Metainebleige Labs (MSL). These include all our foundations, products and fair teams, as well as a new laboratory focused on developing the next generation of our models.

Alexander Van joined Meta to fulfill the duties of our chief employee of the II and leading MSL. Alex worked together for several years, and I consider it the most impressive founder of our generation. It has a clear sense of historical importance of the warden, and as co -founder and the CEO he created Scaleai in a fast -wide campaign involved in the development of virtually all leading models in the industry.

Nat Friedman also joined Meta to collaborate with Alex to head MSL, leading our work on AI products and applied research. NAT will work with the horse to determine its role that goes forward. He ruled GitHub to Microsoft, and recently ruled one of the leading AI investment firms. NAT worked in our Meta Consistory Group during the last year, so it already feels good our roadmap and what we need to do.

We also have several strong new team members who have joined or joined over the past few weeks that I am also happy to share:

• Trapit Bansal -Penner RL on a chain of thought and co-founder of Openai Openai models.
• Shuchao Bi -In GPT-4O and O4-Mini voice mode. Previously, he headed the multimodal after training at Openai.
• Huiwen Chang -The GPT-4O image generation and previously invented Maskgit and Muse Text-Image architecture on Google Research
• Jan lin -Al help to create O3/O4-Mini, GPT-4O, GPT-4.1, GPT-4.5, 4O-IMAGEN and the operator’s reflection stack.
• Joel Bar – The conclusion is anthropic. Earlier in Meta for 11 years on HHVM, Hack, Flow, Redex, performance tools and machine learning.
• Jack re -The -lasting technology training for Gemini and reasoning for Gemini 2.5. The headed Gopher and Chinchilla Early LLM effort in Deepmind.
• Hongyu Ren -GPT-4O, 4O-Mini, O1-Mini, O3-Mini, O3 and O4-Mini. Previously, the lead group for after workout at Openai.
• Johan Shalkwik – Former Google Fellow, an early Sesame and technical presenter for Mayan.
• PEI SUN -In the workout, coding and reflection for twins in Google Deepmind. Previously, the last two generations of Waymo perception models have been created.
• Jiahui yu -The O3, O4-Mini, GPT-4.1 and GPT-4O. Previously, he headed the perception team in Openai and jointly headed the multimodal in Gemini.
• Shengjia Zhao -Chatgpt, GPT-4, all mini-models, 4.1 and O3. Previously, LED synthetic data in Openai.

I’m delighted with the progress we have planned for Llama 4.1 and 4.2. These Meta AI Power models, used by more than 1 billion monthly assets in our applications and an increasing number of meta agents that help improve our products and technology. We strive to continue to build these models.

In parallel, we will start researching the next generation of models to get to the border next year. I have spent the last few months, meeting with the main people in Meta, other laboratories II, and promising startups to gather a founding group for these small talents. We still form this group, and we ask several people through AI Org to join this laboratory.

Meta has a unique position to ensure overseeing the world. We have a strong business that supports construction much more computing than smaller laboratories. We have a deeper experience of building and growing products that reach billions of people. We pioneer and lead the category AI and stretchers that grow very fast. And our company structure allows us to move with much more conviction and courage. I am optimistic that this new talent tributary and a parallel approach to the development of the model will make us fulfill the promise of personal supertelicles for everyone.

We have even more wonderful people at all levels that join these efforts in the coming weeks, so be aware. I am glad to immerse myself and get started.

Watch: Costs on Meta’s talent.