Head over to the Omnilingual demo to explore the languages in the dataset: https://
aidemos.atmeta.com/omnilingualasr
/language-globe
…
@aiatmeta
-
Explore Omnilingual ASR Demo and Dataset Languages
By
–
-
Scaling TorchTitan with SkyPilot Community Tutorials
By
–
More ways to scale TorchTitan with SkyPilot great to see the community expanding the ecosystem with new tutorials like this one!
-
Meta FAIR Introduces Code World Model for Advanced Code Generation Research
By
–
New from Meta FAIR: Code World Model (CWM), a 32B-parameter research model designed to explore how world models can transform code generation and reasoning about code.
— AI at Meta (@AIatMeta) 24 septembre 2025
We believe in advancing research in world modeling and are sharing CWM under a research license to help empower… pic.twitter.com/kwykTJu1GFNew from Meta FAIR: Code World Model (CWM), a 32B-parameter research model designed to explore how world models can transform code generation and reasoning about code. We believe in advancing research in world modeling and are sharing CWM under a research license to help empower
-
Meta Connect 2025 Keynote Livestream: AI Wearables Future
By
–
Meta Connect 2025 is here! Watch the keynote livestream tonight @ 5 pm PT to see the future of AI wearables and beyond: https://
meta.com/connect/ -
Meta releases DINOv3 computer vision model under commercial license
By
–
To help foster innovation and collaboration in the computer vision community, we’re releasing DINOv3 under a commercial license with a full suite of pre-trained backbones, adapters, training and evaluation code, and (much!) more. Find them here: https://
github.com/facebookresear
ch/dinov3
… -
DINOv3 Models Now Available in Hugging Face Transformers
By
–
We're excited to share that we have Day-0 support in Hugging Face Transformers for DINOv3 so people can easily leverage the full family of models. Find out more on @huggingface here:
-
DINOv3: State-of-the-Art Self-Supervised Vision Model
By
–
Introducing DINOv3: a state-of-the-art computer vision model trained with self-supervised learning (SSL) that produces powerful, high-resolution image features. For the first time, a single frozen vision backbone outperforms specialized solutions on multiple long-standing dense… pic.twitter.com/nwS3zFCaaN
— AI at Meta (@AIatMeta) 14 août 2025Introducing DINOv3: a state-of-the-art computer vision model trained with self-supervised learning (SSL) that produces powerful, high-resolution image features. For the first time, a single frozen vision backbone outperforms specialized solutions on multiple long-standing dense
-
DINOv3: Self-Supervised Learning for Large-Scale Vision Models
By
–
A few highlights of DINOv3 👇
— AI at Meta (@AIatMeta) 14 août 2025
1️⃣SSL enables 1.7B-image, 7B-param training without labels, supporting annotation-scarce scenarios including satellite imagery
2️⃣Produces excellent high-resolution features and state-of-the art performance on dense prediction tasks
3️⃣Diverse… pic.twitter.com/6glMC1S3KWA few highlights of DINOv3 SSL enables 1.7B-image, 7B-param training without labels, supporting annotation-scarce scenarios including satellite imagery
Produces excellent high-resolution features and state-of-the art performance on dense prediction tasks
Diverse -
Meta FAIR Wins Algonauts 2025 with TRIBE Brain Modeling
By
–
🏆 We're thrilled to announce that Meta FAIR’s Brain & AI team won 1st place at the prestigious Algonauts 2025 brain modeling competition.
— AI at Meta (@AIatMeta) 11 août 2025
Their 1B parameter model, TRIBE (Trimodal Brain Encoder), is the first deep neural network trained to predict brain responses to stimuli… pic.twitter.com/IeX5gPd8GzWe're thrilled to announce that Meta FAIR’s Brain & AI team won 1st place at the prestigious Algonauts 2025 brain modeling competition. Their 1B parameter model, TRIBE (Trimodal Brain Encoder), is the first deep neural network trained to predict brain responses to stimuli
-
Meta Releases OMC25 Dataset with 25 Million Molecular Crystal Structures
By
–
We’re also releasing the Open Molecular Crystals (OMC25) dataset, comprised of 25 million structures, that was created to enable the FastCSP workflow. Read the paper: https://
ai.meta.com/research/publi
cations/open-molecular-crystals-2025-omc25-dataset-and-models/
… Find the dataset: https://
huggingface.co/facebook/OMC25 We’re excited to share these advances,