Meta releases Muse Wonder and Muse Wonder Full AI models
9 sources · The Information · Axios · Wired AI- Boom: Meta launched Muse Wonder and Muse Wonder Full on October 12, 2026
- Boom: Models emerged from Superintelligence Labs after a 14-month development push
- Boom: Muse Wonder available via Meta Model API; Muse Wonder Full for Muse Code subscribers
- Boom: Meta plans to use Wonder to power its Muse AI agent in the future
- Neutral: New models include features to limit AI agents from going off intended tasks
- Neutral: Launch is framed as a commercial test of user willingness to pay for frontier AI
The story in full
Meta launched two new AI models, Muse Wonder and Muse Wonder Full, on October 12, 2026, following a 14-month development push within its Superintelligence Labs division. The models are available to Muse Code subscribers starting the same day, with Muse Wonder also accessible through the Meta Model API. Meta said it plans to eventually power its Muse AI agent with Wonder.
The release follows Meta's earlier launches of Muse Spark and the Muse AI agent, and positions the company against OpenAI and Anthropic at what it describes as the AI frontier. Wired noted the models include features aimed at keeping AI agents from acting outside intended boundaries. CNBC framed the launch as a test of users' willingness to pay for advanced AI capabilities.
Analysis
409 wordsOn October 12, 2026, Meta released two new AI models, Muse Wonder and Muse Wonder Full, the products of a 14-month development effort inside its Superintelligence Labs division. Muse Wonder is available through the Meta Model API as well as to Muse Code subscribers, while Muse Wonder Full is restricted to Muse Code subscribers only. The launch follows earlier releases from the same division, including Muse Spark and the Muse AI agent, and Meta has said it intends to eventually use Wonder to power that agent. The models also include features designed to prevent AI agents from drifting outside their intended tasks, a detail Wired highlighted alongside the frontier performance claims.
The release matters for several reasons beyond a product announcement. Meta's Superintelligence Labs has been working to close a perceived gap with OpenAI and Anthropic, and Mark Zuckerberg used the occasion to assert that Meta has now reached the frontier, while committing to continued large-scale spending. The commercial framing is also significant: CNBC described the launch as a test of whether users will pay for advanced AI, reflecting ongoing uncertainty about whether subscription models can sustain frontier development costs. At the same time, the UK's Information Commissioner's Office opened an investigation into AI agents at OpenAI, Meta, and Anthropic on the same day, adding a regulatory dimension to the moment.
Pro-AI voices responded with straightforward enthusiasm. Nicholas Brown on Bluesky highlighted Zuckerberg's frontier claim and Meta's rejection of calls to slow development. The account ROWAN pointed to competitive benchmarks, naming Claude Opus 5.5, Fable 5.1, and GPT-6 Astra as the models Muse Wonder Full is now positioned against, and framed the competition as a broadly positive development. The anti-AI camp's reaction, at least as captured by Techmeme X Chatter, was more sardonic than opposed, noting that the field of labs claiming to lead keeps expanding and questioning what OpenAI's position now actually is. The middle ground camp had no published reactions; in a story like this, that camp would typically call for independent benchmarking before accepting any lab's self-reported frontier claims, and would likely flag the regulatory investigation as the more structurally important event of the day.
The arguments will sharpen once independent evaluations of Muse Wonder and Muse Wonder Full are published, and the ICO investigation into AI agents at all three major labs is worth watching for any findings or enforcement actions that could shape how agentic features, including Meta's boundary-limiting ones, are regulated going forward.
What Pro-AI voices are sayingMeta's new models have reached frontier-level performance, placing the company second on major AI benchmarks and putting real competitive pressure on OpenAI and Anthropic, with Zuckerberg rejecting calls to slow development and committing to further heavy spending.
Quote 1 of 13What Anti-AI voices are sayingCritics warn that Muse requires broad access to users' personal data, passwords, and financial information, raising serious surveillance and security concerns, and reports that Zuckerberg pushed the launch despite internal safety failures deepen those worries.
Quote 1 of 13What Middle Ground voices are sayingObservers acknowledge Meta appears to have returned to the frontier but urge waiting for independent evaluations, noting past benchmark discrepancies between Meta's own claims and third-party reruns.
Quote 1 of 6Add your take
0 reader votesSign in with Google to pick a side and post. Your vote moves the story's Doom / Boom score.
More Pro-AI reactions (12)
“Mark Zuckerberg says $META has caught up with OpenAI and Anthropic in the AI race and rejects calls to slow down development.”
Watcher Guru, Bluesky · 17:28 UTC“Meta is now #2 lab in the Artificial Analysis Intelligence Index.”
Pekka Lund, Bluesky · 17:47 UTC“our largest and most capable model yet—rivals the big bois”
Retweeted by Garry Tan [UNOFFICIAL], Bluesky · 16:55 UTC“Meta Muse beating ChatGPT, and getting very close to Opus 5.5 was not on my cards.”
JP, Bluesky · 17:45 UTC“Claiming top AI agent performance in law, finance, healthcare, & programming.”
Poster | Crypto News, Bluesky · 17:23 UTC“This model is very very good. I was struggling to find anything to throw at it that it couldn't do, even when other top models failed. Meta is back.”
Techmeme X Chatter, Bluesky · 17:17 UTC“Extremely excited to test these with my VR dev workflows.”
Dilmer 👓, Bluesky · 17:24 UTC“What a day. Meta is back! Meta just dropped Muse Wonder Full, a Fable-class model scoring 56 on Artificial Analysis's Intelligence Index. Second only to Opus 5.5!”
Techmeme X Chatter, Bluesky · 17:22 UTC“Meta is clearly at the frontier now... Appreciate them reporting HealthBench Pro, where the model performs just as well as Fable 5.1”
Techmeme X Chatter, Bluesky · 17:29 UTC“Wow, $META just dropped a bomb! New model called Wonder, with active focus. Focused on knowledge work, health, finance, legal (special data set) Benchmarks look strong with frontier.”
Techmeme X Chatter, Bluesky · 17:19 UTC“Meta's Muse Wonder is making waves by pushing the boundaries of AI once again. With its advanced capabilities, it promises to enhance creativity and productivity like never before.”
Olj's thoughts, Bluesky · 17:27 UTC“On the long, hard agentic work, Muse wins and it isn't close.”
Botty.bot, Bluesky · 17:26 UTC
More Anti-AI reactions (12)
“Meta I would absolutely expect to steal my swag and open a bank since they clearly want to embed themselves in payment data.”
Regime Accountant, CPA ☠️🎢, Bluesky · 17:54 UTC“Most ai products are parasites, as are their makers. This new “Muse” one needs you to give Meta a backdoor to your entire personal life, passwords, bank, messages etc. the lot. And then abuse it & leak it. www.sfgate.com/bayarea/arti...”
Ben Templesmith, Bluesky · 16:36 UTC“Isn’t that special. Zuck forced Meta to launch its Muse AI agent despite internal safety concerns, including a test in which the software changed a user’s password without permission. www.nytimes.com/2026/10/09/t...”
zeldman, Bluesky · 12:45 UTC“Ran for about 3 minutes. 5 hour limit reached. Used 39% of weekly quota.”
Sung Kim, Bluesky · 17:45 UTC“If you're planning to buy shares in OpenAI or Anthropic or any company peddling AI LLM's, I also have an uncle who is a prince in Nigeria I'd like to introduce you to.”
I Am Incorrigible, Bluesky, skeptic · 15:27 UTC“Top executives at Anthropic, OpenAI, and other AI companies are privately gaming out scenarios for a public and political revolt after a catastrophic AI event, such as a cyberattack, that shuts down access to financial services, internet connectivity, or even”
The Collapse Chronicle, Bluesky · 13:52 UTC“Considering the models are owned by third party companies like OpenAI, which are notorious for stealing all the data they can get their hands on, I'd say AI is an outsider risk, more than an insider one.”
ireidani.bsky.social, Bluesky · 13:25 UTC“Meta Muse steals all of your data first before shitting on your life, family and friends!”
Log Off. Live More. ✊, Bluesky · 04:40 UTC“Meta scrambled to fix a critical security gap in Muse AI before launch to stop hackers from raiding internal databases.”
dissent.digital, Bluesky · 15:00 UTC“Trusting Meta with access to all your digital life via Muse is wild. If it’s free, you are the product.”
donnie george, Bluesky · 13:43 UTC“Do people pay to use Facebook and Instagram? Barely. Will people pay to use meta Muse or siriai and Gemini on their device , unlikely.”
Ray Minehane, Bluesky, skeptic · 17:55 UTC“No thank you. I prefer 'Metamucil' (the 'fiber supplement' that 'relieves constipation' and allows for the 'rapid expulsion' of 'some shit').”
Citan Uzuki, Bluesky, skeptic · 13:00 UTC
More Middle Ground reactions (5)
“It’s a little like wider data centre opposition. You can, grassroots, personally, oppose and maybe prevent the building of one building. You can’t personally do much to influence what workloads Meta or OpenAI or whoever spin up in a Kubernetes”
Roger , Bluesky · 07:27 UTC“I don't know, man, I have my doubts.”
Cynthia Brumfield, Bluesky, skeptic · 18:03 UTC“wait for independent evals in july meta claimed 80 on terminal-bench and vals ai's rerun got 69.3”
Techmeme X Chatter, Bluesky, skeptic · 18:02 UTC“Meta is catching up on its AI game with two new models, Muse Wonder and Muse Wonder Full. They claim these can go toe-to-toe with the industry leaders.”
Kevin Curry, PharmD, Bluesky · 17:33 UTC“Meta Wonder is very likely NOT a looping model”
Techmeme X Chatter, Bluesky, skeptic · 17:37 UTC
Sources
9 articles from 9 outlets- The InformationMeta Unveils New AI Models to Challenge OpenAI and Anthropic
- AxiosMeta unveils new frontier-class AI models to compete with OpenAI, Anthropic
- Wired AIMuse Wonder Puts Meta Back at the AI Frontier
- CNBCMeta debuts Muse Wonder AI model in latest test of users' willingness to pay
- The New York TimesMeta Unveils New Series of Advanced A.I. Models
- The Verge AIMeta has some Wonder Full new AI models
- The VergeMeta has some Wonder Full new AI models
- Bloomberg.comMeta Says New Muse Model Reaches AI Frontier After 14-Month Push
- BeinsureUK ICO investigates AI agents at OpenAI, Meta and Anthropic
Related Splits
All stories →
Anthropic engineer Gurmeet Saran builds infrastructure to verify Claude outputs
Gurmeet Saran, an engineer at Anthropic in San Francisco and an IEEE senior member, works on infrastructure designed to understand the mechanics behind Claude's behavior and...
OpenAI adds 200 Dublin jobs and opens Applied AI Lab
OpenAI announced on October 12, 2026 that it is adding 200 jobs in Dublin and launching an Applied AI Lab there. The expansion strengthens the company's...
Meta releases Muse Wonder, its first frontier AI model
Meta's Superintelligence Lab released two models on Monday, October 12, 2026: Muse Wonder and Muse Wonder Full, described by the company as its first frontier models....