Following the success of its ICML 2024 tutorial, we're now releasing Part 4 of Physics of Language Models—a research effort from FAIR at Meta to uncover universal laws of AI through controlled, scientific experimentation. In Part 4, we challenge today's architecture design
@aiatmeta
-
Collaborative Reasoner: Framework for Social Agents and Human Partnership
By
–
Introducing Collaborative Reasoner: a framework to improve collaborative reasoning in language models.
— AI at Meta (@AIatMeta) 12 mai 2025
Collaborative Reasoner paves the way for developing social agents that can partner with humans and other agents.
Read the research paper and download the code.… pic.twitter.com/btqd8RKcidIntroducing Collaborative Reasoner: a framework to improve collaborative reasoning in language models.
Collaborative Reasoner paves the way for developing social agents that can partner with humans and other agents. Read the research paper and download the code. -
Meta Releases 8B Dynamic Byte Latent Transformer Model Weights
By
–
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.
— AI at Meta (@AIatMeta) 12 mai 2025
Learn more about how Dynamic Byte Latent… pic.twitter.com/1XUtTpq95SWe’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. Learn more about how Dynamic Byte Latent
-
Meta Locate 3D: Advanced 3D Object Localization Model for Robots
By
–
Introducing Meta Locate 3D: a model for accurate object localization in 3D environments.
— AI at Meta (@AIatMeta) 8 mai 2025
Learn how Meta Locate 3D can help robots accurately understand their surroundings and interact more naturally with humans.
You can download the model and dataset, read our research paper,… pic.twitter.com/LP3OBzSF2PIntroducing Meta Locate 3D: a model for accurate object localization in 3D environments. Learn how Meta Locate 3D can help robots accurately understand their surroundings and interact more naturally with humans. You can download the model and dataset, read our research paper,
-
Meta Launches PLM: Open Vision-Language Model for Computer Vision
By
–
Introducing Meta Perception Language Model (PLM): an open & reproducible vision-language model tackling challenging visual tasks.
— AI at Meta (@AIatMeta) 7 mai 2025
Learn more about how PLM can help the open source community build more capable computer vision systems.
Read the research paper, and download the… pic.twitter.com/lcYFZ8EnkrIntroducing Meta Perception Language Model (PLM): an open & reproducible vision-language model tackling challenging visual tasks. Learn more about how PLM can help the open source community build more capable computer vision systems. Read the research paper, and download the
-
Meta Perception Encoder: Advanced Vision Model for Image and Video Tasks
By
–
Introducing Meta Perception Encoder: a vision encoder setting new standards in image & video tasks. It excels in zero-shot classification & retrieval, surpassing existing models.
— AI at Meta (@AIatMeta) 6 mai 2025
Learn more about Meta Perception Encoder, read the research paper, and download the code and dataset… pic.twitter.com/EQVkSVZwgYIntroducing Meta Perception Encoder: a vision encoder setting new standards in image & video tasks. It excels in zero-shot classification & retrieval, surpassing existing models. Learn more about Meta Perception Encoder, read the research paper, and download the code and dataset
-
Meta SAM 2.1 Powers Cutouts Feature in New Edits App
By
–
Curious about the tech behind Edits? Our latest blog post explores how Meta Segment Anything Model (SAM) 2.1 is enabling the Cutouts feature in the new Edits app. Learn more:
-
Solo Tech Receives Llama Impact Grant for Rural AI Support
By
–
Meet Solo Tech, one of the 10 international recipients of the second Llama Impact Grants.
— AI at Meta (@AIatMeta) 30 avril 2025
Solo Tech uses Llama to offer offline, multilingual AI support for underserved rural communities with limited internet access. This grant will help them to equip 50 rural centers with AI… pic.twitter.com/6YflcOlE7jMeet Solo Tech, one of the 10 international recipients of the second Llama Impact Grants. Solo Tech uses Llama to offer offline, multilingual AI support for underserved rural communities with limited internet access. This grant will help them to equip 50 rural centers with AI
-
Llama Stack Partners Enable Modular Development Flexibility
By
–
Cox also gave a special shoutout to our partners, who’ve helped make it possible to get started building with Llama right off the bat—regardless of your development environment. Because the Llama stack is modular, it’s easy to plug and play and switch between component parts.
-
Llama Reaches 1.2 Billion Downloads Milestone
By
–
We announced Llama’s first 1 billion downloads last month, and we’re updating that number to 1.2 billion downloads today. “And what’s cool if you look at Hugging Face where the downloads are happening is that most of these are actually Llama derivatives. We have thousands of