Meta Said to Be Planning Major Push into AI Superintelligence

Determined to rebound from the Llama 4 Behemoth delay, Meta Platforms is swinging for the fences in its artificial intelligence efforts, with CEO Mark Zuckerberg personally supervising creation of a new AI lab focused on superintelligence. Meta has reportedly reached a deal to pay up to $15 billion for a stake in a startup called Scale AI that would bring its 28-year-old co-founder and CEO Alexandr Wang in-house to lead the effort. Other leading Scale AI staff would also come aboard, and Meta is said to be recruiting researchers from OpenAI and Google by dangling “seven- to nine-figure compensation packages,” reports indicate. Continue reading Meta Said to Be Planning Major Push into AI Superintelligence

Meta Restructures AI Ops in Wake of Llama 4 Behemoth Delay

Meta Platforms has restructured its artificial intelligence operations, creating two separate units — an AI Products team headed by Connor Hayes and the AGI Foundations unit jointly run by Ahmad Al-Dahle and Amir Frenkel. The move comes as Meta is delayed getting its latest foundation model Llama 4 Behemoth to market and has also seen several key AI engineers depart. The idea is that dividing a single large organization into two smaller divisions will speed product development and add flexibility as Meta faces tough competition from top companies such as Google and OpenAI, as well as social companies X and TikTok. Continue reading Meta Restructures AI Ops in Wake of Llama 4 Behemoth Delay

Meta Unveils Multimodal Llama 4 Models, Previews Behemoth

Meta Platforms has released its first Llama 4 models, a multimodal trio that ranges from the foundational Behemoth to tiny Scout, with Maverick in between. With 16 experts and only 17B active parameters (the number used per task), Llama Scout is “more powerful than all previous generation Llama models, while fitting in a single Nvidia H100 GPU,” according to Meta. Maverick, with 17B active parameters and 128 experts, is touted as beating GPT-4o and Gemini 2.0 Flash across various benchmarks, “while achieving comparable results to the new DeepSeek v3 on reasoning and coding with less than half the active parameters.” Continue reading Meta Unveils Multimodal Llama 4 Models, Previews Behemoth