MetaInfrastructure· 5 days agoMeta announced MetaRoCE, a new RDMA transport protocol designed for AI workloads on commodity Ethernet, and is releasing its specification, reference software implementation, and compliance test…
MetaInfrastructure· 6 days ago Meta announces MTIA 300, its first training chip with built-in NIC chiplets and communication-offloading engines, designed to optimize communication for training recommendation models, achieving 3.9x…
MetaAvailability change· 18 days ago WhatsApp is introducing Scam Alert, an optional on-device ML feature that detects scam messages without sending any message content to servers, using a transparent model that can be independently…
MetaModel update· 24 days ago Meta introduces a multi-stage sequence model and dense tokenization with target-aware attention to scale sequence learning for ads ranking, achieving lifts in conversions and clicks.
MetaModel update· 26 days ago Meta has improved the training efficiency of its Generative Ads Recommendation Model (GEM) by doubling end-to-end training efficiency to 20–25% Model FLOPs Utilization (MFU) while scaling training…
MetaResearch paper· 15 Jul Meta announces a research area called Hierarchical Interest Representation, an upstream representation layer that learns unified embeddings for ads entities (users, advertisers, products, services)…
MetaCapability change· 13 Jul Meta used sched_ext to build a custom scheduling policy for the Ads service, resulting in a 28% reduction in ads retrieval stage tail latency, 3.28 MW power savings, and a 1.1% increase in ads ranked.
MetaInfrastructure· 1 Jul Meta describes the evolution of its BLOB-storage architecture to support AI workloads, focusing on maximizing GPU utilization and research velocity, with a migration from legacy HDD-based to…
Meta is celebrating 10 years as a sponsor of the Python Software Foundation.
Meta presents a case study on its hybrid approach to privacy-aware infrastructure (PAI) for asset classification, combining LLM-based reasoning with deterministic, human-reviewed rules to handle…
Meta engineers developed ultra-narrow steel-can batteries for smart glasses, achieving widths as narrow as 7mm, and improved capacity across generations, including a 248 mAh cell for display glasses.
Meta describes its multi-year effort to adopt the AV1 video codec for real-time communication, covering codec selection, device eligibility, rate control, error resilience, and the challenges of…
MetaInfrastructure· 3 Jun Meta introduces Instantaneous PowerLoss Storm, a new testing paradigm within its infrastructure for handling and mitigating instant or zero-notice power loss in data centers, built from the ground up…
MetaNew model· 29 Apr Meta released the Llama-Guard-4-12B model on Hugging Face, a safety-oriented multimodal model for image-text-to-text tasks using the transformers library.
MetaModel update· 29 Apr Meta released Llama-Prompt-Guard-2-86M, an 86-million-parameter text classification model based on deberta-v2, hosted on Hugging Face under an 'other' license.
MetaNew model· 29 Apr Meta released a new text-classification model called Llama-Prompt-Guard-2-22M on Hugging Face, designed for safety-related tasks and associated with the Llama 4 family.
MetaNew model· 9 Apr Meta released the Llama-4-Maverick-17B-128E model, an image-text-to-text multimodal model based on transformers, using PyTorch, supporting multiple languages.
MetaOpen-weight release· 9 Apr Meta has released the Llama-4-Scout-17B-16E model on Hugging Face, an image-text-to-text model using the transformers library, with a license marked as 'other'.
MetaNew model· 5 Apr Meta released the Llama-4-Maverick-17B-128E-Original model on Hugging Face.
MetaNew model· 5 Apr Meta released the Llama-4-Scout-17B-16E-Original model on Hugging Face on April 3, 2025.
MetaOpen-weight release· 5 Apr Meta released the Llama-4-Maverick-17B-128E-Instruct-FP8-Original model on Hugging Face.
MetaModel update· 5 Apr Meta released the Llama-4-Scout-17B-16E-Instruct-Original model on Hugging Face, a fine-tuned version of the Llama-4-Scout-17B-16E base model, with license 'other'.
This is a Hugging Face repository for the meta-llama/Llama-3.2-90B-Vision-Instruct model, an image-text-to-text model using the transformers library, created on 2024-09-19 and last modified on…
businessinsider.com,
Gizmodo,
PYMNTS+1 more