What Is Meta Superintelligence Labs?

Meta Superintelligence Labs (MSL) is Meta's umbrella artificial intelligence (AI) organization, led by Chief AI Officer Alexandr Wang.

At Layer3Labs, we build custom workflow automations, so we track who ships the models those automations can run on.

Formed on June 30, 2025, MSL builds the Muse model family and the Muse personal-agent app, with Nat Friedman as head of product and Shengjia Zhao as chief scientist.

What Meta Superintelligence Labs Is

Meta Superintelligence Labs is Meta's primary artificial intelligence division responsible for training the Muse model family and building consumer-facing software agents.

On June 30, 2025, Meta put all its AI teams under MSL, as reported by TechCrunch. Meta's Muse Spark announcement describes the lab's goal as "personal superintelligence."

• Current role: Central AI research, foundation model development, and consumer agent group within Meta.
• Parent organization: Meta Platforms, Inc.
• Primary leadership: Chief AI Officer Alexandr Wang, Head of Product Nat Friedman, and Chief Scientist Shengjia Zhao.
• Core product lines: Muse Spark multimodal models, Muse Glimmer open-weight model, Muse Code, and the Muse personal agent app.
• Recent activity: Launch of Muse Spark 1.3, the Muse personal-agent app, and Muse Gadgets in September and October 2026.

Formation and Executive Leadership

Meta brought Alexandr Wang into the organization as chief AI officer through a $14.3 billion investment in Scale AI for a 49% equity stake. Meta named former GitHub chief executive officer (CEO) Nat Friedman the same day to oversee its AI products and applied research.

Daniel Gross left Safe Superintelligence on June 29, 2025, to join Meta's new lab, according to Implicator. TechCrunch reported that early engineering additions included principal researcher Pei Sun from Google DeepMind and Joel Pobar from Anthropic.

On July 25, 2025, Meta appointed Shengjia Zhao, a co-creator of ChatGPT at OpenAI, as Chief Scientist of Meta Superintelligence Labs, as covered by TechCrunch. Mark Zuckerberg stated publicly that Zhao co-founded the lab and had served as its lead scientist from day one, per Social Samosa. Zhao's OpenAI work is covered in our profile of Shengjia Zhao, and other MSL leaders appear in our list of AI people to watch.

Internal Groups at Meta Superintelligence Labs

In August 2025, Meta Superintelligence Labs structured its operations into four distinct operating divisions, according to Wikipedia.

The four branches include TBD Lab, Fundamental AI Research (FAIR), Products and Applied Research, and MSL Infra. Wang leads TBD Lab, which works on large language models (LLMs). Friedman leads Products and Applied Research, and Aparna Ramani leads MSL Infra.

On October 22, 2025, Meta cut roughly 600 artificial intelligence jobs across FAIR, product AI, and infrastructure teams, as reported by ABC7 News. Chief AI Officer Alexandr Wang sent the memo announcing the cuts, and TBD Lab was not affected. A month later, on November 20, 2025, Yann LeCun left Meta to found Advanced Machine Intelligence Labs, per Wikipedia.

Talent Recruitment and Infrastructure at Meta Superintelligence Labs

Meta Superintelligence Labs expanded its research roster by recruiting personnel from competing artificial intelligence laboratories including OpenAI, Google DeepMind, and Anthropic, as reported by TechCrunch.

According to Wikipedia, Mark Zuckerberg offered select staff at competitor laboratories compensation packages valued between US$1 million and US$100 million. No named outlet has published MSL's total spending or headcount as of October 2026.

Meta's Prometheus data center was planned to deliver 1 gigawatt of computing power by 2026, according to TechCrunch.

The Muse Model Lineage from Spark to Glimmer

Meta Superintelligence Labs released its first model, Muse Spark, on April 8, 2026, presenting it in an official blog post as the first deliverable from a ground-up overhaul of Meta's AI efforts.

Muse Spark was designed as a natively multimodal reasoning model supporting tool use and multi-agent coordination, distributed on the web at meta.ai and in the Meta AI app.

On July 7, 2026, Meta launched Muse Image, the division's first image-generation model, integrated directly into Meta AI, per Meta's announcement. MSL then published Muse Spark 1.1 on July 9, 2026, and Muse Spark 1.2 on August 5, 2026, releasing the latter alongside Muse Code, a terminal-based coding agent described on Wikipedia.

• Muse Spark launched April 8, 2026: Natively multimodal reasoning system featuring tool use and agent orchestration.
• Muse Image launched July 7, 2026: MSL's first image-generation model, available in Meta AI.
• Muse Spark 1.1 launched July 9, 2026: Point release of Muse Spark.
• Muse Spark 1.2 and Muse Code launched August 5, 2026: Muse Code is a terminal-based coding agent.
• Muse Glimmer launched August 2026: Open-weight 30-billion-parameter dense model distributed under an Apache 2.0 license.
• Muse Spark 1.3 launched September 2, 2026: Latent Space reported it ranked #3 on the Artificial Analysis index.

Consumer Applications and Hardware Releases

Meta Superintelligence Labs expanded from foundation model distribution to consumer client software on September 8, 2026, by launching Muse, a standalone personal AI agent app reported by The National.

During the product release, Mark Zuckerberg addressed consumer permission boundaries, stating: "You choose which apps and services Muse has access to and you can disconnect them at any time." Yahoo Finance reported that Muse recorded 2.8 million downloads across the United States and Canada during its first 12 days, and that Meta added about $234 billion in market value over five days. Later, Entrepreneur, in an October 5, 2026 summary of a Wall Street Journal profile, reported that Muse accumulated 5 million United States downloads within 22 days.

On September 22, 2026, Nat Friedman stated on X that Muse was "definitely heavily inspired as a product by OpenClaw" but said it was "built from scratch," as documented by TechCrunch.

On October 2, 2026, the Products and Applied Research group entered local hardware with the release of Muse Gadgets, providing open-source firmware and a Linux software development kit (SDK), per TechCrunch. Meta simultaneously unveiled Muse Home Link, a physical universal serial bus (USB-C) dongle, manufacturing 5,000 units to give free to Muse subscribers.

Muse App Versus Muse Glimmer for Business Use

For a business, MSL's releases differ in where the model runs. Muse Glimmer's weights are open under an Apache 2.0 license, and Wikipedia says it can run entirely offline on a single 24 GB consumer GPU. A team that cannot send data to a consumer service such as the Muse app can therefore run Glimmer on its own hardware.