Community mirror of ai.meta.com/blog + official Engineering at Meta RSS + meta-llama & facebook on Hugging Face.
12 updates · page 1 of 1
Meta engineers describe building the Forum app by bringing an AI agent into the development cycle, including codebase changes, frameworks, and internal tooling.
Why it matters: Engineering teams can learn how Meta adapted its process to build Forum with agentic AI.
Meta describes scaling GEM, its ads recommendation foundation model, with more compute and next-generation hardware across its ads ecosystem.
Why it matters: Ads engineers can follow how Meta plans to extend GEM beyond the recommendation model it introduced last year.
At the 2026 Open Compute Project Global Summit, Meta shared IcePack, its first liquid-cooled network rack, the 102.4T IceCube switch, Flattened Planar Fabric, and the open-source MetaRoCE transport.
Why it matters: Network builders can adopt Meta's open rack, switch, fabric, and transport designs when connecting GPUs.
Meta built a system that improves ads recommendations by using idle inference capacity, with two elasticity levers so that work can expand without compromising baseline performance.
Why it matters: Ads ranking can use spare inference capacity while the baseline recommendation service stays intact.
meta-encoder is a feature-extraction model released under the Apache-2.0 license.
Why it matters: Developers can use meta-encoder for feature extraction under an Apache-2.0 license.
We believe glasses are the best form factor for having AI help throughout your day.
Why it matters: A new product capability you may be able to use right away, without code changes.
We’ve built an AI agent that acts as a secondary expert for a given domain, making deep specialist knowledge readily available and preserved for anyone in an organization to access, share, and build upon.
Why it matters: Developer tooling release — update for the latest fixes, features, and compatibility.
Training and serving frontier AI models depends on fast, reliable networks that move data between GPUs without wasting compute cycles.
Why it matters: A new model option to evaluate — compare quality, price, and context length against what you use today.
MTIA 300 is the first of Meta’s family of in-house training and inference accelerators optimized for training ranking and recommendation models.
Why it matters: A new model option to evaluate — compare quality, price, and context length against what you use today.
WhatsApp is committed to helping people stay safe while protecting the privacy of their messages.
Why it matters: A new product capability you may be able to use right away, without code changes.
Official model update from Meta AI (Meta AI Blog).
Why it matters: A new model option to evaluate — compare quality, price, and context length against what you use today.
Official model update from Meta AI (Meta AI Blog).
Why it matters: A new model option to evaluate — compare quality, price, and context length against what you use today.