According to AI at Meta, the Muse Glimmer 30B open-weight model runs at practical speeds on a single consumer GPU at 4-bit precision and leads peers in tool use and multi-step planning.

According to AI at Meta on X, Meta trained its Muse models, including Muse Wonder, on the Prometheus supercluster, reporting 85%+ pretraining goodput, 90%+ RL goodput, and about 47K GPU hours mean time between interruptions.

According to The Rundown AI, Meta released the West’s strongest open weight model as new factory-floor AI use cases, ChatGPT tool workflows, and a Harvard study on AI’s gender divide shape today’s AI landscape.

According to TIME, Meta's Omnilingual ASR suite was named a Best Inventions of 2026 for delivering automatic speech recognition across 1,600+ languages, including 500 low-coverage languages, as highlighted by AI at Meta.

According to TIME, Meta's Omnilingual ASR suite was named a Best Invention of 2026 for delivering speech recognition across 1,600+ languages, including 500 low-coverage languages, as highlighted by AI at Meta.

According to OpenClaw on X, a weekly roundup highlights new personal AI agent updates from OpenAI, SpaceXSI, Meta, Google, Anthropic, OpenClaw, Nous Research, Perplexity, Manus, Genspark, Lindy, Underdog, MacPaw, and Persona, signaling rapid ecosystem growth.

According to a post by Boris Cherny citing Gray Swan AI on X, Anthropic, Meta, Google DeepMind, and Reflection are sponsoring a $300,000 Indirect Prompt Injection red teaming competition to harden Claude and agentic systems.

According to The Rundown AI, Meta released Muse Spark Open, rated 39 on Artificial Analysis’ Intelligence Index, while Nvidia launched Nemotron 3.5 Super VL, claimed 7x smaller and 5x faster than peers its size.

According to AI at Meta on X, Meta released Muse Spark Open, a US open-weight model for coding and agentic work with 816B total and 42B active parameters, a 1M-token context window, and native multimodal support across text, image, video, and audio.

According to AI at Meta via its researcher blog, the Muse Wonder and Muse Wonder Full models achieve state of the art performance on personal and professional agentic tasks.

According to AI at Meta, a human-led scaffold using Muse Spark 1.3 and Muse Wonder analyzed 5M de-identified EHRs to speed hypothesis generation and produced 10 manuscripts.

According to AI at Meta on X, the team highlights a 14 month journey from rebuilding its AI foundation to pushing the frontier, signaling rapid capability progress.

According to The Rundown AI, Meta released Muse Wonder, with Wonder Full scoring 56 on Artificial Analysis’ Intelligence Index and leading Opus 5.5 on DeepSWE, SWE-Marathon, and MemoryBench; it is available now via Meta Model API and Muse Code.

According to AI at Meta via X, Meta introduced Muse Wonder and Muse Wonder Full frontier models, with Muse Wonder available via the Meta Model API and Muse Code, and Muse Wonder Full available to Muse Code subscribers.

According to ylecun, cheap distillation vs costly training will push frontier foundation models toward free and open access.

According to AIatMeta, mathematicians used Muse Spark 1.1 and 1.2 to co-author six papers solving open problems via meta.ai chat, with peer review.

According to TheRundownAI, Google’s Gemini 4 Argon, Meta’s Muse Desktop setup, and US White House AI initiatives headline today’s practical AI moves.

According to AIatMeta, Meta rolled out Muse Spark, Image, Video, Code, Glimmer, Spark 1.1–1.3, Meta Model API, and Muse, accelerating AI productization.

According to ylecun, Meta’s Galactica faced backlash in 2022 while ChatGPT was later praised, highlighting double standards and missed research value.

According to @CNBC, Wall Street raised views on Nvidia, Apple, Tesla, Meta and AMD, highlighting AI demand, margins and cloud growth, per CNBC.