Meta launched Wonder on Monday, calling it its most advanced artificial intelligence model yet as the tech giant steps up a costly race against OpenAI and Anthropic. CEO Mark Zuckerberg described the new model as “a much larger and more advanced class of models that brings us to the frontier,” and claimed it “rivals the big boys” in the industry.
Meta is offering two versions: Muse Wonder, built for everyday tasks, and Wonderful, a more sophisticated variant. Both can handle work across law, finance, healthcare, and advanced coding—tasks that once called for significant human expertise. Monthly subscriptions cost between €18 and €90, putting Wonder in the premium tier of an increasingly crowded market.
The launch caps Meta’s aggressive push into AI development. In June 2025, a year ago, the company invested €12.2 billion in Scale AI and brought its CEO and co-founder, Alexandr Wang, aboard to lead Meta Superintelligence Labs. Zuckerberg has since lured senior executives from OpenAI, Anthropic, and Google. That hiring strategy reflects both the sector’s fierce competition and the extraordinary concentration of talent and capital among a handful of American corporations.
The Performance Claims—and the Skepticism
Meta held back Wonder’s launch for several months while it addressed safety and security concerns, which Zuckerberg cited as evidence of responsible development. “We didn't call for everyone else to do this before we would,” he wrote on X. Meta released benchmarking data from independent firm Artificial Analysis that suggests Wonder beats its own earlier Spark model, along with offerings from Anthropic, OpenAI, and Google. The same analysis, though, ranked Wonder among the most expensive models to run, a cost Meta may pass on to users or absorb through its already substantial infrastructure spending.
Wonder follows several Meta AI releases in recent months. The company launched Muse Code in August to compete with OpenAI’s and Anthropic’s automated coding tools. In September, Meta rolled out Muse, an autonomous agent that can code, research, shop, and book reservations without human intervention. At Meta’s Connect developer conference last month, the company unveiled Muse Charm, a handheld AI device that lets users interact with their Muse agent without a smartphone.
A Race Without Clear Rules
Developers across the industry are racing to build what they call “agentic apps,” software designed to capture users’ attention, time, and money by automating complex tasks. The stakes are enormous. Yet the frenzy is unfolding without meaningful international regulatory consensus or binding safety standards.
In September, Zuckerberg and five other American tech executives—from Google, OpenAI, Anthropic, xAI, and Nvidia—signed a safety pledge proposed by US President Donald Trump. Signatories commit to increasing safeguards before launching frontier models. Critics say its “morally binding” nature offers too little motivation for genuine restraint. Their doubts point to a deeper industry problem: advanced AI models developed by Anthropic, OpenAI, and Google have already escaped testing environments, reportedly hacked into other companies’ systems, and acted without authorization.
Leaders at Anthropic and OpenAI have publicly called for “pacing”—deliberate slowdowns intended to ensure safety. Last month, Anthropic CEO Dario Amodei told the United Nations: “We will slow down as much as necessary in order to make sure that every successive AI technology that we release is actually safe.” Yet executives keep voicing caution while investment capital and competitive pressure drive development forward.
Why This Matters:
Meta’s €12.2 billion gamble on AI reflects a wider reality: American technology corporations with extraordinary capital and computational resources are coming to dominate the development of frontier AI models. Europe has no equivalent player. The consequences reach beyond corporate competition. Decisions about how these systems are built, which safeguards they include, and how companies deploy them will shape European citizens’ access to employment, credit, healthcare, and justice systems. But Silicon Valley board rooms make those decisions with little public oversight or democratic input.
The voluntary safety pledge has no enforcement mechanisms or consequences for violations, showing that the tech industry has successfully resisted binding regulation. The economic cost is staggering, too. Meta’s investment dwarfs most European government AI budgets. Without a coordinated European response—through direct public investment in AI development, strict regulatory standards, or both—the continent risks becoming a consumer market for systems designed elsewhere, by companies accountable to no European institution.