META RELEASES MUSE WONDER AI MODELS TO COMPETE WITH RIVALS

Meta's Superintelligence Lab released two new premium artificial intelligence models, Muse Wonder and Muse Wonder Full, on Monday, describing them as the company's first attempt at a frontier-level AI system. Meta called the models its "most capable... to date," saying they can carry out agentic tasks including processing insurance claims, explaining lease agreements to renters and building working software applications. The models build on Muse Spark, a lighter AI model Meta launched earlier this year that was designed for low-latency performance. According to Meta, the Wonder models are larger than Muse Spark, with higher benchmark scores and improved results on complex tasks such as coding. Muse Wonder is being made available first to software developers who pay for Meta's Model API.

Meta said the launch follows billions of dollars invested in AI development over the past year and reflects chief executive Mark Zuckerberg's stated ambition to compete at what he has called the "AI frontier," a space Meta says has so far been dominated by Anthropic, OpenAI and Google. Developers can access Muse Wonder through a Power Plan priced at $20 a month or a Maximum Plan priced at $100 a month, Meta said, adding that the model would "soon power experiences in Muse" under separate per-million-token pricing. Meta said Muse Wonder expands on Muse Spark's capabilities with greater processing power and improved performance in law, finance and health-related applications.

The launch comes amid continuing debate over AI safety, with some industry figures, including Anthropic chief executive Dario Amodei and OpenAI's Sam Altman, calling publicly for AI development to slow down. Zuckerberg told the Financial Times this week that he is "much closer" to the view that AI can be engineered with conventional guardrails and built responsibly, rather than treated as a technology that will act unpredictably on its own. Katelyn Chedraoui, a reporter covering artificial intelligence, said Meta's frontier-level launch arrives "amid serious safety concerns in the AI industry" and that safeguards would be essential to any new release. Meta has not said when wider public access to the Wonder models will follow the initial developer-focused rollout.