AI at Meta

📣 Blackwell sets a new inference speed world record — A single NVIDIA DGX B200 server with eight #NVIDIABlackwell GPUs can generate over 1,000 tokens per second (TPS) per user on the Llama 4 Maverick model, the largest and most powerful model in the AI at Meta Llama 4 collection. ⚡⏱️ Additionally, a system with eight Blackwell GPUs can also deliver up to 72,000 tokens/second in a maximum throughput scenario. 🏆 Blackwell is the first platform to achieve this model performance, demonstrating how it delivers the best combination of throughput, output speed and accuracy for LLM token generation. 🏎️🏁 See our tech blog for details ➡️ https://nvda.ws/4kadNB2

Meta FAIR and Rothschild Foundation Hospital present a groundbreaking study mapping how language representations emerge in the brain, revealing striking parallels with large language models (LLMs). 📄 Read the full research paper: ➡️ https://go.fb.me/hfh9i5] This research offers unprecedented insights into the neural development of language, showing how AI models like wav2vec 2.0 and Llama 4 mirror the brain's language processing.

🚀Applications are open for the Llama Startup Program!🚀 We're thrilled to announce the Llama Startup Program, a new initiative designed to empower early-stage startups to innovate and build generative AI applications with Llama. Why Join the Llama Startup Program? ☑️Cloud reimbursements to offset costs. ☑️Hands-on technical support from Llama experts to help you leverage Llama’s capabilities. ☑️A vibrant community and direct access to resources to help your startup thrive. Who Should Apply? Any incorporated, U.S.-based early-stage startup with less than $10 million in funding and at least one developer on staff. Applications are open until May 30, 2025, at 6 p.m. PT Learn more and apply today! ➡️https://go.fb.me/8rn4dh

Introducing Adjoint Sampling, a new learning algorithm that trains generative models based on scalar rewards. Read the research paper ➡️ https://go.fb.me/4mk563 Download the model ➡️ https://go.fb.me/9g4u67 Download the code and benchmark ➡️ https://go.fb.me/g75cpt Based on theoretical foundations developed by FAIR, Adjoint Sampling leads to a highly scalable practical algorithm, and can become the foundation for further research into highly scalable sampling methods.

Exciting news coming out of Microsoft Build: Coming soon, the Llama herd of models will be direct first-party offerings in Azure AI Foundry, hosted and sold directly by Microsoft—with all the SLAs Azure customers expect from any Microsoft product. We’re thrilled to make it even easier for enterprises to build with Llama! #MSBuild 🦙

We’ve released Open Molecules 2025 (OMol25), a new Density Functional Theory (DFT) dataset for molecular chemistry, and Meta's Universal Model for Atoms (UMA), a machine learning interatomic potential. Get started: Download the OMol25 dataset and models: ➡️  https://go.fb.me/m71gmt Read the OMol25 paper: ➡️ https://go.fb.me/tdmssq Download the UMA model: ➡️ https://go.fb.me/kre3aq Read the UMA paper:➡️ https://go.fb.me/dywy19 What it is: These tools will accelerate molecular and materials discovery, unlocking new possibilities for innovation and impact. OMol25 is the largest and most diverse dataset of high accuracy quantum chemistry calculations for biomolecules, metal complexes, and electrolytes. UMA is trained on over 30 billion atoms to provide more accurate predictions and improved understanding of molecular behavior.

Announcing the newest releases from Meta FAIR. We’re releasing new groundbreaking models, benchmarks, and datasets that will transform the way researchers approach molecular property prediction, language processing, and neuroscience. Read more ➡️ https://go.fb.me/ylnz3m Our latest research artifacts include: 1️⃣  Open Molecules 2025 (OMol25): A dataset for molecular discovery with simulations of large atomic systems. 2️⃣ Universal Model for Atoms: A machine learning interatomic potential for modeling atom interactions across a wide range of materials and molecules. 3️⃣ Adjoint Sampling: A scalable algorithm for training generative models based on scalar rewards. 4️⃣ FAIR and the Rothschild Foundation Hospital partnered on a large-scale study that reveals striking parallels between language development in humans and LLMs.

CATransformers is a carbon-driven neural architecture and system hardware co-design framework. Using CATransformers, we discover greener CLIP models that achieve an average of 9.1% reduction potential in total lifecycle carbon emissions while maintaining accuracy (or increasing accuracy) and latency. This research is the first to look into carbon-driven neural architecture and system hardware co-design. It is enabled by a first-of-its-kind architectural carbon modeling tool – ACT, which we developed at FAIR. Check out: Our paper ➡️ https://go.fb.me/nedmh0; code repository ➡️ https://go.fb.me/7jcp6q; and the additional carbon design tools and research artifacts in Sustainable AI ➡️https://go.fb.me/csee32

Introducing Collaborative Reasoner: a framework to improve collaborative reasoning in language models. Download the Collaborative Reasoner code ➡️ https://go.fb.me/6935da Download the MATRIX code ➡️ https://go.fb.me/c098yx Read the research paper ➡️ https://go.fb.me/1zytab Learn how Collaborative Reasoner paves the way for developing social agents that can partner with humans and other agents.

We’re releasing model weights for our 8B- parameter Dynamic Byte Latent Transformer, an alternative to traditional tokenization methods with the potential to redefine the standards for language model efficiency and reliability. Download the model ➡️ https://go.fb.me/utq1ml Download the code ➡️ https://go.fb.me/d4xe6v Read the research paper ➡️ https://go.fb.me/8usw6a Learn more about this significant advancement in byte-level language model architectures.