Meta has unveiled its top-tier artificial intelligence (AI) model series, positioning itself to rival OpenAI and Anthropic.
According to a Bloomberg report on the 12th (local time), Meta introduced the flagship AI models 'Muse Wonder' and 'Muse Wonder Full.' Internally codenamed 'Watermelon,' these models represent the highest-performing AI developed by Meta to date. The company emphasized that Muse Wonder is comparable to OpenAI's 'ChatGPT' and Anthropic's 'Claude.'
Mark Zuckerberg, Meta's CEO, shared on social media, “If 'Muse Spark' was our first model family, 'Muse Wonder' is a much larger and more advanced model that propels us into the cutting edge of technology.”
Meta stated that both models demonstrate top-tier performance in AI agent tasks, including coding and specialized fields like law, finance, and healthcare. Muse Wonder was developed by Meta's Superintelligence Lab (MSL) using the company's proprietary data center infrastructure, 'Prometheus.'
Alexander Wang, Meta's Chief AI Officer (CAIO), told Bloomberg, “Meta Muse is a state-of-the-art AI model, achieving the highest performance levels among existing models, especially in personal and professional AI agent applications.”
Muse Wonder scored 52 points on the AAII, ranking 7th alongside models like Anthropic's 'Claude Pable 5.1,' OpenAI's 'GPT-6 Astra,' and Google's Gemini 4 'Argon' (each scoring 53). 'Muse Wonder Full' achieved 56 points, placing it in a tie for 2nd behind Anthropic's 'Claude Opus 5.5' (58 points).
Notably, the model incorporates 'Active Memory,' which retains conversation history, preferences, and habits, and 'Active Focus,' which attempts alternative approaches to complete tasks when errors occur. Enhanced privacy protection and cyberattack defense features were also added.
Muse Wonder is available via developer APIs and the AI coding tool 'Muse Code.' Muse Wonder Full will be offered to paid subscribers of Muse Code. Meta plans to integrate the new models into its upcoming personal AI agent app, 'Muse.'
