Navigation
Meta AI (FAIR)
OrganizationAlso known as: Meta AI, FAIR, Facebook AI Research, Meta Fundamental AI Research
Revenue
$201.0 billion
as of 2025
Headcount
78,800
as of Dec 2025
Founded Date
Dec 2013
Key People
4YL
Yann LeCunFounder
Chief AI Scientist
Dec 2013 – present
Founded FAIR in 2013; 2018 Turing Award laureate for deep learning; vocal critic of AI x-risk concerns
JP
Joelle Pineau
VP of AI Research
Jan 2017 – present
Leads FAIR; McGill professor; advocates for reproducible research and open science
AA
Ahmad Al Dahle
VP of Generative AI
Jan 2023 – present
Leads Meta's Generative AI group responsible for Llama models
MZ
Mark Zuckerberg
CEO of Meta Platforms
Feb 2004 – present
Drives overall AI strategy for Meta; committed to open-source AI approach
Model Releases
4Apr 2025
Llama 4
Next-generation open-weight models with mixture-of-experts architecture
Apr 2024
LLaMA 3
Major upgrade (8B, 70B params); 405B released July 2024
Jul 2023
LLaMA 2
Open-weight release (7B-70B params) with commercial license
Feb 2023
LLaMA
First LLaMA model family (7B-65B params); initially restricted release, later leaked
Products
2PyTorchSep 2016
Open-source deep learning framework; became de facto standard for AI research
pytorch.orgMeta AI AssistantSep 2023
AI assistant integrated across Meta apps (WhatsApp, Instagram, Messenger, Facebook)
ai.meta.comAll Facts
Financial
Annual Cash Burn$70 billionJan 2026▶
| As Of | Value | Source | Fact ID |
|---|---|---|---|
| Jan 2026 | $70 billion | cnbc.com | f_lCAFEOqqgQ |
Headcount78,800Dec 20252 pts▶
| As Of | Value | Source | Fact ID |
|---|---|---|---|
| Dec 2025 | 78,800 | — | f_F2Y0SdI52g |
| Dec 2024 | 1,500 | — | f_adIN3DEEYw |
Infrastructure Investment$125 billion20263 pts▶
| As Of | Value | Source | Fact ID |
|---|---|---|---|
| 2026 | $125 billion | about.fb.com | f_Yc5dGApsJw |
| Oct 2025 | $27 billion | fortune.com | f_l8ZXhqm3Mw |
| 2025 | $69 billion | about.fb.com | f_4I3Dq2OGcQ |
Revenue$201.0 billion2025▶
| As Of | Value | Source | Fact ID |
|---|---|---|---|
| 2025 | $201.0 billion | — | f_TWxhXOjkvg |
Products & Usage
User Count1 billionApr 20253 pts▶
| As Of | Value | Source | Fact ID |
|---|---|---|---|
| Apr 2025 | 1 billion | medium.com | f_hpCI8c51YQ |
| Mar 2025 | 1 billion | resourcera.com | f_ynNWjLA9og |
| Mar 2025 | 630 million | resourcera.com | f_Mb5ygFtjbA |
Organization
Founded DateDec 2013—▶
| As Of | Value | Source | Fact ID |
|---|---|---|---|
| — | Dec 2013 | ai.meta.com | f_S2gCtemHdA |
HeadquartersMenlo Park, CA—▶
| As Of | Value | Source | Fact ID |
|---|---|---|---|
| — | Menlo Park, CA | ai.meta.com | f_S0d9WF0E8g |
Legal StructureDivision of Meta Platforms, Inc.—▶
| As Of | Value | Source | Fact ID |
|---|---|---|---|
| — | Division of Meta Platforms, Inc. | ai.meta.com | f_Hh20Dvc90g |
People
Founded BycMbVUVK29Q—▶
| As Of | Value | Source | Fact ID |
|---|---|---|---|
| — | Yann LeCun | ai.meta.com | f_lhQD26Idvw |
General
Websitehttps://ai.meta.com/—▶
| As Of | Value | Source | Fact ID |
|---|---|---|---|
| — | https://ai.meta.com/ | wikidata.org | jHzNZtapXA |
Research Areas
4| Name | Description | Team Size | Started | Key Publication | Notes |
|---|---|---|---|---|---|
| Self-Supervised Learning | Learning representations from unlabeled data; foundation of modern LLM training | — | Jan 2014 | — | Core research philosophy under Yann LeCun; underpins Meta's approach to foundation models |
| Computer Vision | Image recognition, object detection, and visual understanding | — | Dec 2013 | arxiv.org | Segment Anything Model (SAM) released 2023; DINOv2 self-supervised vision transformer |
| Open-Source AI | Developing and releasing AI models and tools under open licenses | — | Jan 2015 | ai.meta.com | Meta's flagship open-source strategy; PyTorch (2016), Llama series, SAM, and numerous research tools |
| Natural Language Processing | Language understanding, generation, and translation | — | Dec 2013 | — | NLLB (No Language Left Behind) translation for 200+ languages; contributed to multilingual AI research |
Safety Milestones
3| Name | Date | Type | Description | Source | Notes |
|---|---|---|---|---|---|
| Llama Guard | Dec 2023 | safety-eval | Open-source LLM-based safeguard model for content moderation | ai.meta.com | Released with Llama model family; classifies prompt and response safety |
| Responsible Use Guide | Jul 2023 | policy-update | Guidelines for responsible deployment of Llama models | ai.meta.com | Acceptable use policy accompanying Llama 2 open-weight release |
| Purple Llama | Dec 2023 | safety-eval | Open-source tools for evaluating and improving LLM safety including CyberSecEval | ai.meta.com | Collaborative project for AI safety evaluations; includes Llama Guard and CyberSecEval benchmarks |
▶Internal Metadata
| ID | tt0f5PYDCw |
| Stable ID | tt0f5PYDCw |
| Numeric ID | E549 |
| Type | organization |
| YAML Source | packages/kb/data/things/tt0f5PYDCw.yaml |
| Facts | 16 structured (18 total) |
| Records | 17 in 5 collections |