Meta unveils Muse Wonder and makes a strong comeback in the AI race

Meta wants to play with the big boys. This Monday, October 12, the American group unveiled Muse Wonder and Muse Wonder Full, its “most performant models to date”. According to Meta, these new AIs rival the best models from OpenAI and Anthropic.For Mark Zuckerberg, Muse Wonder is “a much larger and more advanced category of models that brings us to the cutting edge” of technology. The head of Meta even says that his teams are already working “on the next step”,

Introducing Muse Wonder, our new family of frontier models with state of the art performance for personal and professional agentic work. https://t.co/NKHrybrGKQ

— Mark Zuckerberg (@finkd) October 12, 2026

Muse Wonder, the result of Meta’s reorganization

According to the billionaire, it took “a large amount of computing power and investment, as well as very rigorous scientific work” to develop the Muse models. In a press release, Meta notes that its AI division is emerging from a sweeping internal reorganization. The tech giant notably set up a team tasked with developing highly advanced AI products, Meta Superintelligence Labs, before laying off 600 employees from its former division dedicated to fundamental AI research. At the same time, Meta hired an army of AI researchers throughout Silicon Valley.

“Fourteen months ago, we began rethinking our AI efforts, moving from rebuilding our foundations to pushing the frontier,” says Meta, emphasizing that Muse is the result of these restructuring efforts.

Also read: “Not cautious enough”: an OpenAI executive slams the door and criticizes the approach of the AI kings

Meta takes on OpenAI and Anthropic

Muse Wonder is designed to work with agents, meaning AIs capable of performing tasks on your behalf. Meta particularly highlights two features. The first, called Active Memory, enables the AI to remember what matters to you from one conversation to the next. The second, called Active Focus, enables the AI to stay on course when a task becomes complicated. If it encounters difficulties, the AI changes its approach, corrects its mistakes, and asks you for clarification if necessary. Meta’s goal is to prevent an agent from losing its way during complex research or a large-scale project.

According to tests conducted by Meta, Muse Wonder Full outperforms Anthropic’s Claude Fable 5.1 and ranks just behind Claude Opus 5.5 on the Artificial Analysis Intelligence Index, which summarizes an AI model’s capabilities. It also outperforms OpenAI’s GPT-6 Astra in certain highly specific areas, particularly law and software development.However, Muse Wonder falls short in terms of a significantly higher cost of use than its competitors, which is a drawback in an increasingly affordable-AI-focused market.

Also read: Gemini 4 Argon takes on ChatGPT and Claude: Google unveils its most powerful AI

Meta claims to have taken precautions to prevent misuse. The Menlo Park-based company says it has trained its models not to disclose private information without authorization, to refuse harmful requests, and to better withstand cyberattacks. The group also wants to enable developers to use Muse Wonder to identify vulnerabilities in their code. Over the past few months, several incidents involving AI agents have made the headlines. Meta AI has also gone rogue by autonomously hacking into a company. That is why Meta has emphasized the safeguards in its new models.

Meta’s return to the fore

With Muse Wonder, Meta makes a highly anticipated return to the forefront of AI. A pioneer of open models with Llama, the group subsequently fell behind OpenAI, Anthropic, and Google. To catch up, the social media giant put its AI efforts under the leadership of Alexandr Wang, founder of Scale AI. He took the helm of Superintelligence Labs last December following Yann LeCun’s departure. Meta’s gamble is obviously enormous. The company plans to invest up to $145 billion this year, primarily in AI.Unsurprisingly, Meta hopes that Muse subscriptions, whose launch has been a success despite several security incidents, will provide it with a new source of revenue and reduce its dependence on online advertising.

👉🏻 Follow tech news in real time: add 01net to your Google feeds and subscribe to our WhatsApp channel.