Meta Unveils Muse Wonder: A Paid Shift to Frontier AI

Key points

• Meta released Muse Wonder and Muse Wonder Full on October 12, 2026, positioning them as state-of-the-art models for personal and professional AI agents.
• The models feature 'Active Memory' to retain context across conversations and 'Active Focus' to sustain effort on long-running tasks, including complex coding and domain-specific expertise in law, finance, and health.
• This launch represents a shift from Meta's previous open-source strategy to a paid, closed-source model, with subscription tiers ranging from 20 cents per million tokens to $200 per month.
• The Muse agent, launched in early September 2026, has seen rapid adoption, accounting for 72% of top agent-based internet traffic according to Human Security.
• Meta is investing heavily in infrastructure, with projected data center spending of at least $600 billion by 2028, while CEO Mark Zuckerberg rejects calls for an AI development slowdown.

Background

Meta's AI division underwent a major restructuring in 2025 under chief AI officer Alexandr Wang. The company had previously focused on open-source models but has now pivoted to paid, closed-source technology to compete directly with OpenAI and Anthropic. This move follows the launch of the Muse agent in September 2026, which quickly gained popularity. The broader AI industry has seen increased competition, with OpenAI launching its 'Dots' agents in late September 2026 to rival Meta's offerings.

How outlets are covering it

Meta and its CEO, Mark Zuckerberg, emphasize the models' frontier capabilities and the company's focus on personal empowerment rather than enterprise automation. Zuckerberg told the Financial Times that Meta is 'on the frontier' and rejected calls for an industry-wide AI slowdown, arguing that safety should be managed through individual lab guardrails. In contrast, Gizmodo highlights the significant financial risks, noting Meta's previous $83 billion loss in the metaverse and its current $600 billion data center spending projection. Gizmodo also points to security concerns, reporting that Meta found serious flaws in the product just before launch. Reuters notes that Muse's agentic abilities have sparked concern in sectors like travel and tax preparation, while Citigroup projects $27 billion in annual revenue for Muse by 2030.

Why it matters

The launch of Muse Wonder signals a major shift in Meta's AI strategy, moving from open-source to paid, closed-source frontier technology. This could impact the competitive landscape, as Meta now directly challenges OpenAI and Anthropic in both consumer and enterprise markets. The model's focus on personal agents and its integration with Meta's existing products could reshape how users interact with AI, potentially disrupting traditional services. However, the massive infrastructure spending and security concerns raise questions about the sustainability and safety of Meta's AI ambitions.

What to watch

Meta plans to expand the use of Muse Wonder across its products, including its Muse AI agent. The company will also compete in the enterprise market by selling access to the latest models. A full safety report with additional evaluations is expected soon. The market will watch how the paid subscription model performs and whether Meta can sustain its massive data center investments. Analysts will also monitor the impact on competing AI services and the potential for regulatory scrutiny regarding data privacy and security.