Zuckerberg creates Meta Superintelligence Labs

Background

Before MSL, Meta's AI work was spread across several groups. Its foundations team, product team and the FAIR research lab each had their own roles. Meta had also been building the Llama family of open models, which faced questions in mid-2025 about the future of its largest model. On 14 July 2025 the New York Times reported that MSL leaders had discussed abandoning Llama 4 Behemoth in favour of a closed model.^[2]

What happened

On 30 June 2025 Mark Zuckerberg announced Meta Superintelligence Labs (MSL) in an internal memo.^[1] Alexandr Wang became chief AI officer and leader of MSL. Former GitHub chief executive Nat Friedman co-led the lab, heading AI products and applied research.^[1]

Key details

Structure. MSL combined Meta's foundations, product and FAIR teams with a new lab focused on next-generation models.^[1]

Hires. The memo named new hires from OpenAI, Google DeepMind and Anthropic, including Shengjia Zhao, Jiahui Yu, Shuchao Bi, Hongyu Ren, Jack Rae and Joel Pobar.^[1]

Scale. The memo said Meta AI had more than 1 billion monthly active users. It set a goal of reaching the frontier "in the next year or so".^[1]

Personal superintelligence. On 30 July 2025 Zuckerberg published a letter titled "Personal Superintelligence", which set out MSL's goal of a personal superintelligence for everyone.^[3]

Reorganisation. In August 2025 MSL was reorganised into four groups: TBD Lab, FAIR, Products and Applied Research, and MSL Infra.^[4]

First model. MSL's first model, Muse Spark, was released on 8 April 2026.^[5]

Impact

MSL marked a change in how Meta organised its AI work: research, products and infrastructure were placed under one leadership, with a stated goal of reaching the frontier. The later Muse models, covered in the entries that follow, came from this lab. The reorganisation in August 2025 and the reported discussion of Behemoth both fell within the lab's first months.^[1][2][4]