Meta Unveils Muse AI Models to Compete in Advanced Artificial Intelligence

Image: Illustrative image · Meta Digital Simple Human · Public domain · Source

Meta launches Muse Wonder and Muse Wonder Full AI models targeting complex personal and professional tasks, featuring active memory and improved task handling, directly competing with OpenAI and Anthropic.

Meta Platforms CEO Mark Zuckerberg announced the launch of two new artificial intelligence models, Muse Wonder and Muse Wonder Full, representing a significant advance in the company's AI capabilities. These models are designed to handle complex tasks across personal and professional domains, including programming, document analysis, legal interpretation, personal finance, and health data management.

After investing heavily in computing infrastructure and AI research, Meta says these new models outperform its previous AI solutions. They incorporate features such as Active Memory, which enables the AI to retain relevant user information during ongoing conversations, enhancing context awareness and continuity. Another feature, Active Focus, helps the models manage longer tasks, adjust to new information dynamically, correct errors, and seek user clarification before making critical decisions.

Performance benchmarks indicate that Muse Wonder offers strong results in agent performance, memory, and programming tasks, while the advanced Muse Wonder Full provides better outcomes in handling novel tasks. This positions the models as competitive alternatives to other leading AI chatbots, including Anthropic's Claude Opus 5.5 and Fable 5.1, and OpenAI's GPT-6 Astra, with Meta asserting superiority in certain technical metrics.

Unlike earlier open-source strategies, Meta will distribute Muse Wonder through proprietary development and subscription channels. Pricing reported by the Financial Times suggests basic access starting at $0.20 per million tokens, with the advanced Muse Wonder Full available at about $200 per month. This marks a clear positioning against rival providers, leveraging Meta’s investments and extensive research to capture enterprise AI engagements.

Zuckerberg emphasized that reaching this frontier of AI capability demanded substantial computational resources and continuous research, noting ongoing development of new models and planned further investments. He also downplayed concerns about AI safety, expressing confidence that such technologies can be responsibly designed with standard safeguards to prevent misuse and unauthorized data disclosures.

Meta underlines that Muse models are trained to avoid sharing private information improperly and to reject harmful requests while maintaining abilities useful for defensive cybersecurity tasks. This launch represents a strategic step as Meta seeks to solidify its role in the competitive AI industry landscape, offering advanced AI assistance for both individuals and organizations.

The information is sourced from Mobile World Live and Financial Times reports dated October 2026.