Meta challenges its rivals with the Muse Wonder and Muse Wonder Full AI model duo.

On October 12th, the US-based technology company Meta announced two new artificial intelligence (AI) models called Muse Wonder and Muse Wonder Full, claiming that these models have capabilities comparable to those of OpenAI and Anthropic.

Meta CEO Mark Zuckerberg stated on social media platform X that the Muse Wonder line of models is larger and more advanced than the previously launched Muse Spark line, placing Meta in the ranks of the most advanced AI model developers.

Meta says paying users will soon be able to use two new AI models through Muse, the company's personal AI assistant app.

According to Meta's Superintelligence Labs, Muse Wonder is Meta's largest and most capable model to date, and is strong enough to compete with leading competitors. In a blog post, Meta cites benchmark data published by Artificial Analysis, an independent AI testing and evaluation organization.

According to this data, the new Wonder models performed better than Meta's previous Spark models, and also surpassed several models developed by Anthropic, OpenAI, and Google.

Meta has committed billions of dollars to compete in the global AI development race, despite growing concerns about safety and security. In July, the conglomerate that owns social media networks Facebook and Instagram said its total spending on AI investments this year could reach $145 billion.

Over the past year or so, Zuckerberg has been pushing Meta to compete with the world's leading AI research labs, particularly those of OpenAI and Anthropic. The company aims to attract more customers and diversify its revenue streams, which have long relied heavily on advertising.

In April 2026, Meta launched Muse Spark, an AI model that marked a shift from its previous open-source Llama line. Meta moved towards developing AI models with a new focus, as Llama was considered to be lagging behind its competitors in the market.

By July 2026, the company will introduce Muse Spark 1.1, which it describes as its most capable AI model in practical programming and automated task execution. This is part of Meta's broader goal of developing "personal superintelligence."