Meta Superintelligence Labs
Meta Superintelligence Labs is Meta Platforms' artificial-intelligence division focused on advanced AI research, infrastructure, and consumer products.
Last updated August 31, 2026
AI crawler reads
No AI crawler has read this profile since Oct 28, 2026.
Overview
Meta Superintelligence Labs (MSL) is the artificial-intelligence division of Meta Platforms, established in June 2025 as part of a major effort to consolidate Meta's advanced AI research, model development, infrastructure, and product work. The division was created against the backdrop of intensifying competition among Meta, OpenAI, Google, Anthropic, and other technology companies to develop increasingly capable foundation models and systems associated with artificial general intelligence and artificial superintelligence. MSL grew out of Meta's earlier artificial-intelligence organizations, particularly Facebook Artificial Intelligence Research (FAIR), later renamed Fundamental Artificial Intelligence Research after Facebook became Meta, and the broader Meta AI organization. FAIR had been founded in 2013 and became known for research in computer vision, natural-language processing, self-supervised learning, generative models, translation, and machine learning infrastructure. Its most influential technical contribution was PyTorch, the open-source deep-learning framework released in 2017. PyTorch subsequently became a major tool across academic and commercial AI development. The immediate catalyst for MSL was Meta's effort to strengthen its large-language-model program after the release of Llama 4 in April 2025. Meta began developing a more ambitious model, reportedly codenamed Behemoth, while senior leadership increased its direct involvement in recruiting AI researchers and reorganizing the company's AI activities. Meta also made a reported $14.3 billion investment for a 49% non-voting stake in Scale AI and recruited Scale AI founder Alexandr Wang to lead the new organization as chief AI officer. Former GitHub chief executive Nat Friedman joined to lead product-oriented work, with Daniel Gross later becoming associated with the product leadership structure. MSL incorporates several functions: a team responsible for large-language-model development, FAIR research, products and applied research, and infrastructure engineering. The organization therefore combines long-horizon scientific research with the development of models and services intended for Meta's consumer ecosystem, including the Meta AI assistant. The division's reported Muse model family includes Muse Spark, Muse Image, and Muse Video, although the availability and status of individual models may vary by market and release stage. The organization remains closely tied to Meta's wider AI portfolio. Its predecessors contributed to PyTorch, the Llama series, Meta AI, and specialized research systems such as Galactica. Meta's AI work also depends on substantial computing infrastructure, including Nvidia graphics processors and internally developed acceleration hardware. Through MSL, Meta has sought to make advanced AI a central strategic capability spanning social platforms, messaging, recommendation systems, developer tools, and generative media. MSL's public identity is still developing. It is primarily an internal division rather than a separately incorporated company or consumer-facing standalone brand. Its significance derives from Meta's capital investment, executive recruitment, model roadmap, and attempt to compete at the frontier of AI research while integrating those capabilities into Meta products.
History
Meta Superintelligence Labs is the latest organizational expression of Meta's long-running artificial-intelligence program. Its principal predecessor, Facebook Artificial Intelligence Research, was created in 2013. FAIR developed research programs in computer vision, natural-language processing, machine translation, self-supervised learning, generative adversarial networks, and document understanding. It operated across several international research locations and participated in the 2016 Partnership on AI alongside major technology companies and research organizations. Yann LeCun directed FAIR during its formative period and was succeeded by Jérôme Pesenti in 2018. FAIR's most durable contribution was PyTorch, released in 2017 as an open-source deep-learning framework. PyTorch became widely used in research and production systems, helping establish Meta as an important infrastructure contributor even when its consumer AI products were less prominent. FAIR also developed language and chatbot experiments, including a 2017 project that attracted inaccurate reports that its systems had invented an unintelligible language. Meta explained that the experiment had simply reached its research objective and was discontinued for that reason. Following Facebook's corporate rebranding as Meta in 2021, FAIR was renamed Fundamental Artificial Intelligence Research and operated alongside the newly emphasized Meta AI organization. Meta subsequently developed the Llama family of large language models, releasing successive generations from 2023 through Llama 4 in 2025. The company's release strategy combined research publication, developer access, and varying degrees of openness, making Llama an important part of Meta's effort to compete with proprietary foundation-model providers. The formation of MSL in 2025 followed internal pressure to improve Meta's frontier-model performance. After Llama 4, Meta began work on a more ambitious system reportedly codenamed Behemoth. Concerns about the model's readiness contributed to a delay, while Mark Zuckerberg became more personally involved in recruiting researchers and organizing a new advanced-AI effort. Meta pursued investments and possible acquisitions involving Scale AI, Safe Superintelligence, Thinking Machines Lab, and Perplexity AI. The completed Scale AI transaction gave Meta a substantial non-voting stake and brought Alexandr Wang into the company. On June 30, 2025, Meta announced MSL, placing Meta AI, FAIR, and additional model and product teams within the new structure. Alexandr Wang became chief AI officer, Nat Friedman led product work, and later reporting associated Daniel Gross with the product leadership group. MSL was subsequently divided into four broad areas: a model-development group sometimes referred to as TBD Lab, FAIR research, products and applied research, and MSL Infrastructure. The division's reported Muse family extends its remit beyond text generation to image and video creation. Muse Spark was described as a model powering the Meta AI assistant, while Muse Glimmer was presented as an open-weights model intended to run on personal computers. The reported model roadmap illustrates MSL's goal of combining frontier research with direct deployment in Meta's consumer services. However, model availability, release dates, licensing, and benchmark claims should be checked against primary announcements because the organization and its products are evolving rapidly.
- 2025
Meta Superintelligence Labs is founded
[1]Meta formed MSL to combine advanced AI research, model development, products, and infrastructure.
- 2025
Meta invests in Scale AI
[1]Meta announced an investment of approximately $14.3 billion for a 49% non-voting stake in Scale AI and recruited Alexandr Wang.
- 2023
Meta begins the Llama model series
[1]Meta released the first Llama large language model, beginning a major foundation-model program that later fed into MSL.
- 2021
FAIR is renamed Fundamental AI Research
[1]Following Facebook's rebranding as Meta, FAIR received a new expanded name and continued as part of Meta's AI organization.
- 2017
PyTorch is released
[1]FAIR released PyTorch, which became a widely adopted open-source framework for deep-learning research and deployment.
- 2016
FAIR joins the Partnership on AI
[1]FAIR participated in the formation of an industry and research partnership focused on beneficial and responsible AI development.
- 2013
Facebook establishes FAIR
[1]Facebook created Facebook Artificial Intelligence Research, the principal research predecessor of Meta Superintelligence Labs.
Products and positioning
An advanced-AI research and product organization positioned at the intersection of frontier model development, open and open-weight AI, large-scale computing infrastructure, and integration into Meta's global consumer platforms.
Large language model · 2026
Muse Spark
A Muse-family generative AI model associated with Meta Superintelligence Labs and described as powering the Meta AI assistant. It represents MSL's effort to connect advanced model research with a consumer-facing conversational product.
Generative image model
Muse Image
A reported member of the Muse family focused on image generation. Publicly available technical and commercial details are limited in the supplied reference material.
Generative video model
Muse Video
A reported Muse-family system for generative video. It extends the division's model portfolio beyond text and image generation toward synthetic moving-image creation.
Large language model family · 2023
Llama
Llama is Meta's major large-language-model family and a key predecessor program to MSL. Releases from 2023 through Llama 4 established Meta's strategy of distributing capable foundation models to researchers and developers under varying access and licensing terms.
Machine-learning framework · 2017
PyTorch
Released by FAIR in 2017, PyTorch is an open-source framework for building and training neural networks. It is infrastructure rather than an MSL consumer product, but it remains one of the most significant technical outputs of the division's research lineage.
Controversies
2022
Galactica withdrawal
FAIR made the Galactica scientific-text-generation model publicly available for a short period before withdrawing it after users demonstrated racist, inaccurate, and misleading outputs. The episode became an example of the risks involved in releasing generative models without sufficient safeguards and evaluation.
Brand decisions
2025
Create a dedicated superintelligence organizationStrategyBackground
Meta faced stronger competition in frontier AI and concerns about the performance and readiness of its latest large language models.
Action taken
Meta consolidated Meta AI, FAIR-related research, a new model-development group, product teams, and infrastructure functions under Meta Superintelligence Labs.
Aftermath
The reorganization increased executive attention, accelerated researcher recruitment, and made advanced AI a more distinct strategic priority within Meta.
- OpenAI — Meta's recruiting drive contributed to wider competition among AI companies for senior researchers and technical staff.
- Microsoft — Microsoft chief executive Satya Nadella was reported to have responded to the intensified competition for AI talent.
2025
Investment in Scale AIM&ABackground
Meta sought to expand its access to AI data, evaluation capabilities, and senior technical leadership.
Action taken
Meta announced an approximately $14.3 billion investment for a 49% non-voting stake in Scale AI and recruited Alexandr Wang to lead Meta's advanced AI effort.
Aftermath
The transaction supplied MSL with a prominent AI executive and strengthened Meta's position in the competition for frontier-model talent and infrastructure.
Investment value. $14.3 billion announced investment (2025)
Recent events
- Unconfirmed
Omacom Foundation launches with $23.5 million in pledges and donations[2]
Reported:
The Omacom Foundation was incorporated as a nonprofit to hold trademarks, fund infrastructure, promote the work, and support open-source projects and developers Omarchy depends on. Twelve Founding Patrons each contributed $1 million, corporate patrons pledged $1 million a year for three years or $1.5 million in tokens, and over 800 donors raised over $120,000, bringing total pledges and donations to approximately $23.5 million.
Sources2 sources
Cite this profile: Cite the canonical profile. /brand-wiki/meta-superintelligence-labs · Editorial policy · How profiles are compiled