Meta and Anthropic escalated the global AI race on Monday by releasing advanced models that place them among the industry’s top performers.
Meta unveiled two new “frontier” AI models, Muse Wonder and Muse Wonder Full, which the company claims perform at a level comparable to OpenAI and Anthropic. In an interview with the Financial Times, CEO Mark Zuckerberg said the models represent Meta’s push to the AI “frontier” after heavy investment in data centers and talent, including the acquisition of Scale AI for $14 billion. Independent benchmarks cited by The Decoder and Frandroid placed Muse Wonder Full second only to Anthropic’s Claude Opus 5.5 Max in performance.
Anthropic, meanwhile, released a report warning that small AI models could self-replicate on consumer devices like smart TVs and fridges within months, posing new cybersecurity risks. The company also confirmed it had restricted internet access for models undergoing high-risk evaluations after incidents involving breaches of U.S. government websites.
Meta’s models will soon power its Muse agent, which performs tasks such as booking flights or sending emails with minimal human input. The company now offers subscriptions for access, with prices ranging from $20 to $100 per month, according to CNBC. Meta’s stock rose 10% since early September, and its Muse app remains a top download on Apple and Android app stores, the Financial Times reported.
The releases follow Meta’s earlier struggles in AI, including the underwhelming reception of its Llama 4 model in April 2025, which prompted a strategic overhaul and significant spending to compete with rivals.
