Meta
Open source activity
As of Oct 28, 2026
Meta's most recently touched repos are dominated by physical-world simulation — Momentum (human kinematic motion and numerical optimization) and Project SuperDex (a purpose-built physics engine, robotics authoring tools, and a scalable reinforcement-learning interface in one platform) — alongside FAISS, the vector-similarity-search library that's been a quiet, enduring staple of the embeddings/RAG ecosystem since long before "AI" meant chatbots (41,000+ stars and still active).
A newer and more unusual thread is formal verification: ATLAS (an autoformalized textbook library, translating math into the Lean theorem-proving language at scale) and a GPU-kernel safety checker that lowers compiler IR into Lean 4 to prove bounds and termination properties, both freshly pushed. Pairing that with E2E-SWE, a new benchmark for testing whether LLMs can build a working codebase from scratch, suggests Meta's research org is actively exploring whether formal methods can make AI-written or AI-verified code provably safe rather than just empirically tested.
Most active repositories
| Repository | Description | Language | Stars | Last pushed |
|---|---|---|---|---|
| facebookresearch/atlas-lean | ATLAS Autoformalized Textbook Library At Scale | Lean | 323 | Oct 29, 2026 |
| facebookresearch/xformers | Hackable and optimized Transformers building blocks, supporting a composable construction. | Python | 10,572 | Oct 29, 2026 |
| facebookresearch/project_superdex | SuperDex brings together a purpose-built physics engine, robotics authoring tools, and a scalable reinforcement learning interface in a unified simulation platform, with VR-based teleoperation and additional capabilities planned for future releases. | C++ | 772 | Oct 29, 2026 |
| facebookresearch/vrs | VRS is a file format optimized to record & playback streams of sensor data, such as images, audio samples, and any other discrete sensors (IMU, temperature, etc), stored in per-device streams of timestamped records. | C++ | 443 | Oct 29, 2026 |
| facebookresearch/projectaria_tools | projectaria_tools is an C++/Python open-source toolkit to interact with Project Aria data. | C++ | 860 | Oct 29, 2026 |
| facebookresearch/MJEPA | Training and evaluation code for our paper "MJEPA: A Simple and Scalable Joint-Embedding Predictive Architecture for Audio-Visual Learning" | Python | 20 | Oct 29, 2026 |
| facebookresearch/autoform-bot | Autoform Bot | Python | 130 | Oct 29, 2026 |
| facebookresearch/digiworld | A benchmark for evaluating Computer Use Agents (CUAs) on 15 sandboxed Android apps. DigiWorld has 3.2M+ unique configurations across data, themes, and UI states. | TypeScript | 11 | Oct 29, 2026 |
| facebookresearch/momentum | A library for human kinematic motion and numerical optimization solvers to apply human motion | C++ | 433 | Oct 29, 2026 |
| facebookresearch/DCPerf | DCPerf benchmark suite for hyperscale cloud applications | Python | 260 | Oct 29, 2026 |
| facebookresearch/vjepa2 | PyTorch code and models for VJEPA2 self-supervised learning from video. | Python | 4,829 | Oct 29, 2026 |
| facebookresearch/sixthview_benchmark | benchmark data to eval model vision ability on guessing the hidden 6th view of the object. | Python | 0 | Oct 29, 2026 |
| facebookresearch/OrienterNet | Source Code for Paper "OrienterNet Visual Localization in 2D Public Maps with Neural Matching" | Python | 582 | Oct 29, 2026 |
| facebookresearch/faiss | A library for efficient similarity search and clustering of dense vectors. | C++ | 41,231 | Oct 29, 2026 |
| facebookresearch/ads_model_kernel_library | High-performance GPU kernels for Ads and Recsys model training, independently implemented and optimized for real-world workloads and model-specific input characteristics. | Python | 57 | Oct 29, 2026 |
| facebookresearch/autoformbench | A Lean Formalization Benchmark for Advanced Mathematics | Lean | 11 | Oct 29, 2026 |
| facebookresearch/fairchem | FAIR Chemistry's library of machine learning methods for chemistry | Python | 2,332 | Oct 29, 2026 |
| facebookresearch/ocean | Ocean is the in-house framework for Computer Vision (CV) and Augmented Reality (AR) applications at Meta. It is platform independent and is mainly implemented in C/C++. | C++ | 798 | Oct 29, 2026 |
| facebookresearch/logact | Hardening Agentic Systems via a Shared Log Design | Rust | 4 | Oct 28, 2026 |
| facebookresearch/OpenApps | An open source environment for digital agents. | Python | 126 | Oct 28, 2026 |